Products
AI products organized by the 5-layer stack. Try them, compare them, find what works.
More filtersType, scores & evidence
Selections apply immediately and are saved in the page link.
561 products
≈ Editorial ratings · MethodologyNVIDIA datacenter GPUs across Rubin, Blackwell and Hopper generations
ChatGPT app, GPT models and Codex in one AI family
Claude app, Claude models and Claude Code in one AI family
Google’s AI apps, models, coding tools and enterprise platform in one family
High-bandwidth memory for GPU and AI systems
Stacked high-bandwidth memory for AI accelerators
HBM3E and HBM4 stacked DRAM for AI accelerators
The GitHub of machine learning
Alibaba foundation-model family for language, code and multimodal applications
Coding agents, editor assistance and pull-request review inside GitHub workflows
Open-weight frontier reasoning and agent model family
Voice synthesis, dubbing, music and conversational agents
The AI-first code editor
GPU cloud for model training and inference
Datacenter GPU family spanning MI300, MI350 and MI400 generations
Grok app, models and Grok Build in one AI family
DeepMind's protein structure prediction model
Legal and professional-services AI platform
GPU instances and clusters for AI training and inference
Camera-perception software for automotive ADAS production programs
AI image generation for artists
Customer agents across chat, voice and messaging
Build software with AI
Inference platform for open and fine-tuned models
AI video generation
Enterprise foundation models
ByteDance's enterprise model family on Volcano Engine Ark
Video generation model with audio, references and editing
Managed inference on customer-owned or hosted compute
AI-powered cancer detection
AI-powered enterprise knowledge agent
AI video generation and creative editing suite
Cerebras wafer-scale AI system, from WSE-3 chip to Condor Galaxy clusters
Google Cloud's fifth-generation TPU accelerators — v5e and v5p
Mistral AI's general-purpose language model family
FDA-cleared deep-learning MRI reconstruction.
Huawei's Ascend 910 AI accelerator series — 910B and 910C
Open-source workflow automation and AI agent platform
AI diagnostics for medical imaging
An inference accelerator evaluated as a complete serving system
Kuaishou AI video-generation product
AI answer engine for cited search and research.
Revenue intelligence over recorded sales conversations
AI SRE agents for incidents and production operations
Moonshot AI's assistant and open-weight K-series frontier models
Training data, human feedback and model evaluation
AI virtual metrology for semiconductor fabs
Tencent foundation model family
AI rare-disease whole-exome interpretation.
AI creation and editing inside Canva
Suno's AI music generation product
Rebellions ATOM inference systems from single server to rack-scale POD
AI brand protection against counterfeits and IP infringement
Z.ai (Zhipu) GLM model family for agentic engineering
Meta open-weight models for existing and self-hosted deployments
Korea's sovereign AI foundation model
Photo-based math problem solving and tutoring
Google's Apache-licensed open model family for local and edge AI
De-facto orchestration framework for LLM applications
Groq's LPU inference hardware, from chip to node and rack
Lovable's prompt-to-app builder
Japanese language-model family for API and private deployment
Open-source LLM application development platform
Local and hosted runtime for open models
Dark-web threat intelligence and security analytics
Qwen model-building and application platform
Agent-centered enterprise model platform from Baidu AI Cloud
MiniMax's M-series reasoning and agentic model family
3 kW FuriosaAI inference appliance for agentic systems
Enterprise LLM assistant over internal documents
Operate industrial AI across constrained enterprise environments
Enterprise language models with API and deployment choices
Synthetic avatar video and translation
Lablup's GPU cluster orchestration platform, self-hosted or managed
CXL controller and switch IP for AI infrastructure.
Trace and evaluate agents in actual service workflows
Korean foundation-model program with internal LG deployment and open-weight releases
Synthesia's AI avatar video platform for enterprise communication
Wafer-scale inference cloud for low-latency LLM serving
Developer inference cloud powered by Groq LPU silicon
AI wildfire detection on utility camera networks
3D machine vision and robot-planning suite for industrial automation
Autonomous drone AI for wind, energy, and infrastructure inspection.
Robot controllers and fleet software for industrial mobile robots
AI imaging triage for lung, chest and stroke workflows
AI hospital early-warning system for patient deterioration.
AI-assisted annotation and dataset operations
Infrastructure-based autonomous-vehicle control system.
Ambient clinical documentation inside the EHR
Expressive AI voice and synthetic-actor studio.
Licensable Korean NPU IP for edge-AI chips
AI research agent and cited search engine.
Digital-twin simulation for autonomous driving
Vision AI data management
Desktop workspace and runtime for open models
Korean call, meeting and everyday assistant from SK Telecom
Unified LLM routing and inference marketplace
3D perception software that turns infrastructure LiDAR into AI insights.
AMD's adaptive SoC/FPGA platform for edge AI inference in industrial, automotive, and aerospace deployments.
AI detection of cerebral aneurysms on brain MRA
Industrial vision models connected to inspection operations
Train and deploy visual defect inspection on production lines
Hardware-agnostic robot navigation stack
Enterprise face-verification engine
AI document and drawing workflows for AEC firms
Enterprise generative-AI platform from LG CNS
LLM app and agent builder platform
Swarm light-show drone flown in fleets of thousands
Clinical and research assistant for Tempus reports and records
Conversational AI for NLU, speech, and personalization.
AI customer-support agent inside Channel Talk.
AI business messenger for SMBs.
Demand and materials forecasts for ordering decisions
Chat API platform (7B convos/month)
Omnichannel AI customer service and agent platform
Korean enterprise and public-sector AI chatbot
Edge-AI NPU for industrial vision and robotics.
AI assistant across Meta apps and the web
Decentralized GPU cloud for AI training and inference.
Browser-based AI dental restoration CAD with editable crown designs
AI for renewable-energy site selection and PPA structuring.
Self-hostable Friendli Engine for inference in your own environment.
Managed APIs and dedicated capacity for open models
AMD's open-source GPU computing platform — the software layer enabling AI and HPC workloads on Instinct accelerators.
Tesla's vision-only autonomous driving stack
VESSL's MLOps platform for training, serving, and pipelines.
AI-native risk, lending and service stack for digital banking
Connect enterprise data to supervised agent workflows
Autonomous indoor delivery robot.
Wearable walking-assist robot generating real-world movement data.
Korea's consumer AI platform aggregating frontier models
Amazon's multimodal foundation-model family in Bedrock
Model compression and hardware-aware inference optimization
Find and repair gaps in AI training data
Video understanding API
Baidu's ERNIE model and assistant family
Meta’s Muse model family, separate from Llama
Conversational AI shopping agent for e-commerce.
IONIQ 5 vehicle-platform integration for Waymo One autonomous ride-hailing.
AI fashion image auto-tagging.
Enterprise QA on machine reading comprehension.
AI CRM for D2C merchants on Cafe24 / Naver SmartStore.
Duolingo's AI tier with roleplay, explanations and video call
Korean enterprise generative-AI and RAG platform.
AI development and operations delivered with implementation support
Industrial AI PC with 80 TOPS NPU.
FDA-cleared AI OB/GYN ultrasound — identifies 39 standard views, labels 47 structures, reduces keystrokes 94%.
Face recognition and computer-vision APIs and SDKs
AI detection across Proofpoint's threat protection platform
Identity verification and fraud platform
Tesla's humanoid robot
Open-source desktop assistant for local or remote models
Amazon Bedrock foundation-model family
Cohere's enterprise model family for RAG and tool use
Hosted model APIs for language, video, speech and music
Multimodal AI video-interview assessment
AI software and programming stack for Doosan collaborative robots
Korean deep-learning text-to-speech suite
BGM subscription powered by Pozalabs AI.
Cambricon's datacentre AI accelerators and Neuware toolchain
Classroom Q&A platform with AI clustering and teaching assistance
Figure AI's general-purpose humanoid robot line
Academic research AI for paper-based research.
Vision-AI food scanning and dietary analytics
MFDS-approved AI digital therapeutic for insomnia.
Live multilingual captioning and translation for events.
Media-localization machine translation for film and TV.
Multilingual lip-sync dubbing for video.
Korean/Japanese speech-to-text API
Korean-optimized enterprise speech-to-text engine
Synthetic AI humans for kiosks, video and virtual presenters
Sequence-based AI drug discovery platform
De novo antibody and protein design platform.
Transcriptomics-based AI drug repositioning
AI drug discovery platform.
Enterprise AI services
Baichuan's model API and enterprise agent platform
Mistral AI production platform for building and operating AI systems
Managed AI training capacity on NVIDIA GPU clusters, rented through major clouds
Developer access for StepFun models
Tencent Cloud's end-to-end machine-learning platform
Post-training infrastructure behind a Python API
ByteDance cloud model and agent platform
Z.ai API and model platform
Market intelligence search across filings, transcripts and expert calls
LLM safety and de-identification API.
Spatial AI digital twin platform.
Financial crime and transaction fraud prevention platform
Fraud, compliance and risk platform with device intelligence
AI for financial crime
Behavioural AI against email compromise and account takeover
Continuous blood-pressure monitoring ring.
Embedding, reranking and retrieval tooling for enterprise RAG
Self-learning network anomaly detection and response
Unified data and AI platform
End-to-end data science and AI platform for enterprises
Enterprise AI platform for AutoML, apps and observability
Programmable smart actuators used in robotics research worldwide.
AI across Elastic's search, observability and security stack
AutoML, document AI and open LLMs on one platform
Alibaba's open model hub and Model-as-a-Service community
Developer security scanning with AI code analysis
The agile mobile robot
AI-driven network detection and response
Weaviate's open-source vector database
Cloud security graph with AI posture management
Software-defined vehicle platform for Hyundai Motor Group.
Open-source coding harness with model-provider choice
PCIe 6.4 / CXL 3.2 fabric switch for AI servers.
AI webtoon production studio.
Baidu autonomous-driving and intelligent-vehicle platform
Hanwha Aerospace multipurpose unmanned ground vehicle.
Huawei's Atlas AI compute systems for datacentre and edge
Horizon Robotics' assisted-driving stack on Journey silicon
Huawei's open-source deep-learning framework for Ascend
L4 autonomous-driving software in paid commercial service
Radiology triage platform across clinical specialties
Retinal image analysis for chronic-disease screening
AI 12-lead ECG screening for hidden cardiac disease
FDA Breakthrough–designated brain CT AI for hemorrhage and hematoma — Medicare NTAP reimbursed.
12 FDA-cleared findings from a single chest X-ray — pneumothorax, effusion, vertebral fractures, and more.
First FDA-cleared continuous AI sepsis monitoring — 18% lower mortality when clinicians act on alerts.
10 FDA clearances for stroke imaging triage — mandated across all NHS England and Wales stroke centers.
AI monitoring of pulmonary fibrosis progression from CT — co-primary endpoint in Boehringer Ingelheim Phase 3 trial.
AI vessel suppression for chest CT improving nodule visibility; opportunistic CAC scoring from routine ungated CTs.
AI bone suppression and line/tube confirmation for chest X-ray — improves subtle lung lesion detection.
FDA-cleared digital pathology AI platform for primary cancer diagnosis and pharma drug development.
AI cardiovascular disease risk prediction from a single retinal photograph — MFDS cleared, 8-country international deployment.
Three first-of-kind FDA clearances detecting fatal cardiac conditions invisible to human readers on standard ECG.
FDA-cleared AI stethoscope algorithm detecting low ejection fraction — CMS reimbursed at $128.90/use.
Autonomous diabetic retinopathy screening cleared for two camera manufacturers — Norway's national AI eye screening system.
Non-invasive coronary physiology from CT — replaces diagnostic catheterization for stable chest pain.
AI coronary plaque vulnerability analysis from CTA — identifies rupture-prone plaques; nationally covered by Aetna and Cigna.
FDA-cleared AI for stroke imaging triage.
The first autonomous FDA-cleared AI diagnostic in any medical specialty — diabetic retinopathy without physician review.
FDA-cleared real-time coronary angiography AI — vessel segmentation and lesion quantification in 1–2 seconds.
AI brain MRI for Alzheimer's diagnostics.
AI for digital pathology slide review
Korean humanoid and manipulation robotics platform
RapidAI's imaging platform for stroke and vascular triage
MFDS and FDA-cleared AI echocardiography analysis — 70%+ reporting time reduction; ~3,000 institutions worldwide.
FDA-cleared AI reduces MRI scan time by up to 60% while maintaining diagnostic quality.
FDA-cleared AI enabling 4x faster PET scans at lower radiotracer dose — scanner-agnostic across major OEMs.
AI for precision medicine
AI layer across United Imaging's scanner and radiotherapy fleet
World's first FDA-cleared radiomics AI for lung cancer risk prediction — CPT-reimbursed at $650/procedure.
Care-coordination platform for time-critical imaging findings
Level 4 autonomous-driving stack for shuttles, robotaxis, and AV trucks.
Turn captured objects into editable 3D assets
Hanwha Robotics collaborative robot line.
Big-data credit AI for alt-lending.
Korean lawyer marketplace and legal information portal
ML credit scoring for near-prime borrowers.
AI legal assistant for Korean lawyers and legal teams
World's most dynamic humanoid
Build and operate HyperCLOVA X applications on NAVER Cloud
AI training-data construction service for LLMs.
Autonomous mobile robots and orchestration for warehouse fulfilment
Safe humanoid for human environments
UBTech's humanoid robot line for service and industrial work
Affordable quadruped robot
Unitree's humanoid robot line
FDA-cleared AI for prostate cancer grading.
Rebellions server for efficient large-scale AI inference
Hyundai Motor Group's mobile eccentric droid robotics platform.
Interactive AI avatar interface for kiosks and service environments
Autonomous food-service robot at chain scale
Kia's highway driver-assistance feature set.
AI meeting and call analysis for teams
AI-native entertainment and character-chat platform
Autonomous-driving and vision data labelling
Automotive AI compute platform for driver assistance
4D imaging radar with LiDAR-equivalent perception.
AMD's integrated NPU for consumer and commercial laptops, enabling Copilot+ PC and on-device AI features.
Dexterous robot policies, with deployment still to prove
Korea's first MFDS-approved generative AI radiology report generator — 57 chest X-ray conditions.
Coronary artery calcium scoring from CT
Quantitative lung texture analysis on CT
AI test-prep tutor for TOEIC and TOEFL
Korean collaborative robots with AI vision
Open-weight models for agentic software work
Autonomous defense drone built on Nearthlab's industrial-inspection stack.
Remote identity verification for Korean finance
AI KYC / KYB / AML identity verification.
Medical-grade ring with deep-learning AFib detection.
AI interpretation of satellite imagery
MFDS-approved VR digital therapeutic for stroke.
Retinal-photo AI screening for chronic-disease complications
DEEPX's DX-M on-device inference NPU series — DX-M1 and DX-M2
Korean-language enterprise search, LLM and analytics for closed networks
A pipeline from demonstrations to robot policies
Enterprise AI agent build-and-deploy platform
Deploy and operate agents on AWS
AI video creation and editing tools in CapCut
Captions' AI video creation and editing app
ByteDance AI agent and bot-building platform
Text-based audio and video editing
AI image generation
Optimize and deploy models for target devices
Build internal apps, workflows and AI agents on business data
Adversarial testing for models, retrieval systems and agents
AI upscaling, denoising and restoration for photo and video
Enterprise brand-conditioned content generation
Agent builder for enterprise document workflows
Browser video editor with AI subtitles, avatars and generation
AI video editor that cuts footage by transcript.
Agent building and deployment for Korean consumer AI
AI agents and chatbots across Zapier's automation graph
Phenomics-driven drug discovery platform
AI across Datadog's observability and security platform
AI customer support agents for enterprise service teams
Enterprise AI customer experience agent platform
Enterprise AI contact center
AI triage and autofix over application errors
Predictive analytics and anomaly detection over machine data
AI agents, copilot and quality management in Zendesk
DEEPX high-performance edge AI accelerator
Korean MyData personal-finance AI.
AI insurance coverage and claims automation.
Huawei's compute architecture and toolchain for Ascend silicon
Search, summarisation and review across stored content
1.58-bit quantization for AI models on Arm SoCs.
Serverless GPU inference platform for generative media — image, video, audio, and 3D at scale.
Compiler and runtime for FuriosaAI NPUs
Cloud Hancom Office with AI drafting tier
Kakao's AI platform and business messaging stack
Unified proxy and router across LLM providers
Serverless GPU compute for AI workloads
GPU cluster software and virtualization for AI training
Writing, Q&A and automation across a Notion workspace
Containerized inference microservices for deploying models on NVIDIA GPUs
Open-source LLM inference optimization library for NVIDIA GPUs
Open-source model serving for multi-framework inference
Run and fine-tune open models via API
On-demand and reserved GPU pods for AI training and inference.
Autoscaling serverless GPU endpoints for AI inference.
AI cross-border remittance with FX hedging and fraud detection.
Developer documentation for SiliconFlow model APIs
Hosted model API and MaaS platform
Enterprise private deployment for model-serving workloads
Summarisation, search and workflow automation in Slack
AI co-pilot for Swit work OS.
On-device AI engine (no GPU needed)
Korean enterprise work OS.
Inference, fine-tuning and GPU clusters for open models
VESSL's GPU Liquidity Layer for multi-cloud GPU capacity.
Consumer financial AI assistant built on Ant's finance model
ML-driven drug discovery
MFDS-approved AI for osteoporosis screening from routine chest X-rays — addresses 75% undiagnosis rate pre-fracture.
Baidu AI coding assistant
Continue's open-source AI coding assistant for IDEs
AI assistant and coding agent inside JetBrains IDEs
Test generation and code review agents
Description under review
Tencent Cloud AI coding assistant and IDE
ByteDance's AI development family spanning TRAE IDE and TRAE Work
Z.ai coding and agent product surface
AI character and interactive storytelling service
Japanese-focused language model
Vector search and RAG pipelines for enterprise data
Korea's largest crowdsourced AI training-data platform
Automotive AI accelerator for ADAS and cockpit AI.
Open weights for coding and agentic reasoning
Open-weight reasoning models from OpenAI.
Korean humanoid robot trained on real-world human movement data.
Heterogeneous robot fleet-control platform
AI image creation with a direct editing workflow
AI research over patent, chemical and biomedical literature
Naver's AI-based product recommendation technology for shopping.
AI search, merchandising and recommendations
AI shopping assistant
End-to-end autonomy software for drones and vehicles
AI-backed matching inside Hyperconnect's Azar video-chat service.
Commerce search, personalization and marketing automation
AI video-review collection and merchandising.
Commerce search and discovery optimised for revenue
Personalization and recommendations for commerce experiences
AI across eBay's buying and selling flows
Commerce agents for pricing and product intelligence
JD Cloud enterprise-agent architecture
Software and connected-services platform for Kia PBV vehicles.
AI across Klarna's shopping and merchant surfaces
Game AI for character behaviour and companion NPCs
Meituan foundation-model family
Generative-AI chest X-ray report drafting
Conversational AI shopping agent in the Naver Plus Store app.
Naver shopping membership and commerce surface
Naver's AI-personalized shopping app and commerce surface.
Commerce personalization across search, recommendations and segments
ML-driven Korean lifestyle commerce.
Visual search for fashion e-commerce.
AI assistant for merchants
Unified mobility OS for autonomous Hyundai/Kia vehicles.
WATCHA PEDIA's rating-based recommendation engine.
Chinese social-commerce AI
AI cross-channel ad reporting and optimization.
AI across SMS and email marketing for retail
Korean-tuned enterprise RAG platform.
Predictive targeting and send-time AI in customer engagement
ML-powered native-ad and content-recommendation network.
AI drafting and document assistance for Hancom Office environments
AI across Klaviyo's marketing automation platform
Enterprise AI workspace for search, assistants and agents
Multi-model agent orchestration behind one API
Korean NLP enterprise search engine
Document workspace with source-grounded answers and extraction
FuriosaAI vision accelerator for computer-vision workloads
Xiaomi multimodal models for coding and agents
SK Telecom's open-weight Korean foundation-model series
Ambient clinical documentation and coding
Humanoid robot platform for industrial work
AI logistics management platform
AI cooking robot that replicates chef-grade grilling with real-time sensing.
Open-source ROS reference mobile robot.
Description under review
Description under review
Korean financial data terminal with document and company intelligence
Korea's largest AI robo-advisor by AUM.
S&P Global's AI for financial document extraction and linking
AI-managed ETFs and asset-management infrastructure
Description under review
Z.ai's CogVideoX video-generation model family
MiniMax video creation application
AI creative suite for video, images and advertising production
Video, 3D and neural capture generation
Multimodal video model with native audio
Pika's AI video generation product
AI video-generation platform and API from AIsphere
Shengshu's AI video generation model and product
Alibaba's Wan image and video generation family
Kakao's open-weight Korean foundation model, moving to agents in KakaoTalk
Conversational AI, voice and AI humans for Korean enterprises
LG AI Research's open pathology foundation-model family
AI hiring-match platform for Korean tech talent
Research toolkit for audio generation and compression
Audio AI for spatial sound, loudness and source separation
Creator AI music composition tool.
MiniMax music-generation product line
MiniMax speech model and audio product line
Voice synthesis, conversion, dubbing and speech cleanup
AI music generation with editing controls
Korean AI agent for nuanced search and deep research
AI research and briefing assistant for professionals
AI-generated search briefings embedded in Naver Search
Naver's browser-based AI voice and video dubbing service
Korean STT, transcription, and meeting AI.
Machine translation, writing assistance and speech
Hybrid human + AI translation platform.
AI translation and discovery for Korean webtoons and web novels
AI localization pipeline for subtitles and dubbing
Bimanual robot manipulation foundation model
AI drug discovery with 230M+ data entries.
Life-sciences research model for biology and drug-discovery workflows
Self-serve SaaS AI drug-discovery workbench
AI CGM-based blood-glucose management app
ML and LLM observability with open-source tracing
Evaluation and observability for LLM product teams
Conversation analytics for LLM product teams
AI governance, policy and regulatory readiness
Training-data production and LLM evaluation for Korean AI builders
AI model evaluation and monitoring
Open-source LLM evaluation and red-teaming framework
Evidently's open-source ML and LLM evaluation and monitoring
AI-assisted 3D sketching.
Model monitoring, explainability and LLM guardrails
Evaluation, monitoring and guardrails for LLM applications
Open-source testing and vulnerability scanning for AI models
Guardrails AI's validation framework for LLM output
Great Expectations data-quality validation, open-source and cloud
Runtime guardrails and red-teaming for LLM applications
Open-source LLM engineering and observability platform
LLM observability, tracing, and evaluation platform
Generate, edit and export persistent 3D worlds
3D asset generation from text and images
Generative AI for 3D character motion from text, voice, or video.
AI auto-tracking smartphone camera mount.
AI motion capture in the browser.
Open-source prompt testing and LLM red-teaming
Open-source evaluation framework for RAG systems
Experiment tracking and LLM observability for ML teams
Data and ML monitoring with privacy-preserving telemetry
AI21 hybrid SSM-Transformer model family
ModelBest's efficient multimodal model family for edge deployment
Open-weight Korean LLM trained from scratch.
StepFun's Step foundation-model family
Korean diffusion-transformer LLM.
AI-powered UI generation by Vercel
01.AI open and commercial foundation-model family
Voice AI assistant
Upstage's messenger-based AI assistant with document OCR
ByteDance AI assistant and model-facing consumer app
All-in-one AI workspace and agentic research platform.
SK Telecom's Korean voice-assistant and device ecosystem
iFLYTEK's Spark assistant and model family
Mistral AI assistant and long-horizon work agent, formerly Le Chat
Tencent AI assistant
Framework for programming and optimising LLM pipelines
Connect agents to documents and enterprise data
Framework for customizing and evaluating models on NVIDIA infrastructure
China's largest open-source deep-learning framework and model platform
TypeScript toolkit and gateway for building AI applications
AI loan-prediction and marketplace.
Z.ai vision and image-generation model surface
AI image generation inside a stock-asset library
Persistent agent runtime with reusable skills
Ideogram's AI image generator with reliable in-image text
Real-time generative image and video canvas
Image generation with fine-tuned models for game and product art
Meitu's imaging and video AI across photo, video and design
Photorealistic AI images
Recraft's AI design and vector-graphics generator
Established image-model family for self-hosted and API generation
AI document scanner for mobile.
Meitu AI image creation product
Extensible agent harness and coding CLI
Plan, generate and edit visual content in one studio
Open-weight MoE reasoning model for evaluation and adaptation
BHSN's all-in-one legal AI platform for enterprise legal and business teams.
Korean case-law database with AI search and analysis
AI lesson planning and teaching tools for K-12 educators
Adaptive tutoring system for Chinese K-12 after-school learning
AI avatar video creation platform
Generative virtual-human platform behind Eternity
AI virtual-idol platform for K-pop.
Real-time AI avatar and digital-human studio.
Virtual-human video generation.
Tencent digital-human and avatar product surface
AI across Alibaba's DingTalk workplace suite
AI across ByteDance's Feishu/Lark workplace suite
A compact device for local language-model use
AI video note-taking with transcript Q&A.
AI call-recording and transcription app
Health-behavior AI agent recognized at CES 2026.
Adaptive learning across subjects with teacher analytics
AI coaching and course authoring inside Coursera
AI-assisted coding education and cloud learning platform
AI study tools over Quizlet's study-set corpus
Output-focused K-12 writing and speaking learning
ESL learning with licensed real-world content
Beginner-friendly ESL speaking practice
Ringle's post-lesson AI speech analysis feature.
AI assistant and learning platform over company knowledge
Financial education with market-data-powered AI guidance
AI disease-risk prediction from health-checkup data
Open multimodal models built for customization
Small time-series models for probabilistic forecasting
AI writing workspace connected to research.
Typed decisions and confidence estimates for software
Healthcare-tuned open model family from Google
Open vision-language model family from Google
Therapeutics and drug-discovery open model family from Google
NC AI's game/content foundation-AI platform with a physical-AI pivot
EXAONE scientific discovery model effort
4nm LLM-inference accelerator on LPU architecture
Chat with AI characters
MiniMax AI companion and character product
Brand-conditioned marketing content generation
AI features across the Shopify merchant surface
AI rewriting and writing assistance
Research image-and-text base model with downloadable weights
Early-stage Korean general-purpose humanoid platform.
Product | Company | Layer | Revenue / usage | Product Strength | Product Value | Adoption | Durability | Competitors | |
|---|---|---|---|---|---|---|---|---|---|
NVIDIA datacenter GPUs across Rubin, Blackwell and Hopper generations Sourced | NVIDIA | L1 | 🌐 | Product revenue not established | 92.9 | 96 | 95 | 88 | Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p AWS Trainium |
ChatGPT app, GPT models and Codex in one AI family AI familySourced | OpenAI | L2 | 🌐 | 900M+ WAU · Mar 2026 · ChatGPT app only | 92.2 | 93 | 98.1 | 86 | Claude Gemini |
Claude app, Claude models and Claude Code in one AI family AI familySourced | Anthropic | L2 | 🌐 | Not disclosed · Claude app only | 91.2 | 93 | 94.9 | 86 | ChatGPT Gemini |
Google’s AI apps, models, coding tools and enterprise platform in one family AI familySourced | L2 | 🌐 | 950M MAU · Jul 2026 · Gemini app only | 91.2 | 88 | 98.1 | 88 | ||
High-bandwidth memory for GPU and AI systems PlatformSourced | SK Hynix (SK하이닉스) | L1 | 🇰🇷 | 50% global HBM revenue share · Q2 2026 (est.) | 86.8 | 84 | 95 | 82 | Samsung Electronics (삼성전자) · HBM Memory HBM3E/HBM4 Micron Technology · HBM Memory HBM3E/HBM4 |
Stacked high-bandwidth memory for AI accelerators PlatformSourced | Samsung Electronics (삼성전자) | L1 | 🇰🇷 | 33% global HBM revenue share · Q2 2026 (est.) | 86 | 84 | 92.4 | 82 | SK Hynix (SK하이닉스) · HBM Memory HBM3E/HBM4 Micron Technology · HBM Memory HBM3E/HBM4 |
HBM3E and HBM4 stacked DRAM for AI accelerators PlatformSourced | Micron Technology | L1 | 🌐 | 18% global HBM revenue share · Q2 2026 (est.) | 84.9 | 84 | 88.7 | 82 | Samsung Electronics (삼성전자) · HBM Memory HBM3E/HBM4 SK Hynix (SK하이닉스) · HBM Memory HBM3E/HBM4 |
The GitHub of machine learning PlatformSourced | Hugging Face | L3 | 🌐 | 13M registered accounts in 2025; not active users | 83.9 | 88 | 98.8 | 68 | |
Alibaba foundation-model family for language, code and multimodal applications ModelSourced | Alibaba (阿里巴巴) | L2 | 🇨🇳 | Not disclosed | 81.7 | 87 | 87.2 | 72 | |
Coding agents, editor assistance and pull-request review inside GitHub workflows Agent harnessSourced | Microsoft | L4 | 🌐 | 4.7M paid subscriptions/seats · 2026-01-28 | 81 | 87 | 89.7 | 68 | |
Open-weight frontier reasoning and agent model family ModelSourced | DeepSeek (深度求索) | L2 | 🇨🇳 | Not disclosed | 77.4 | 87 | 80.9 | 66 | |
Voice synthesis, dubbing, music and conversational agents Sourced | ElevenLabs | L5 | 🌐 | $500M+ ARR · 2026-04-30 | 76.8 | 87 | 85.5 | 61 | |
The AI-first code editor Agent harnessSourced | Cursor (Anysphere) | L4 | 🌐 | $3B annualized revenue · 2026-04-30 | 74.6 | 89 | 93.3 | 50 | GitHub Copilot Claude Replit |
GPU cloud for model training and inference PlatformSourced | CoreWeave | L3 | 🌐 | Product revenue not established | 72.2 | 86 | 60 | 73 | - |
Datacenter GPU family spanning MI300, MI350 and MI400 generations Sourced | AMD | L1 | 🌐 | Product revenue not established | 71.6 | 85 | 60 | 72 | Datacenter GPU family Cloud TPU v5e/v5p Intel Gaudi 3 |
Grok app, models and Grok Build in one AI family AI familySourced | SpaceXSI | L2 | 🌐 | Not disclosed · Grok app only | 70 | 78 | 55 | 80 | ChatGPT Claude Gemini |
DeepMind's protein structure prediction model Partly sourced | L5 | 🌐 | Active-user count not verified; see recorded sources | 69.9 | 86 | 72.2 | 55 | ||
Legal and professional-services AI platform AppSourced | Harvey | L5 | 🌐 | $400M+ ARR · 2026-09-09 | 68.9 | 78 | 84 | 50 | |
GPU instances and clusters for AI training and inference PlatformSourced | Lambda | L3 | 🌐 | Product revenue not established | 68.4 | 82 | 60 | 65 | - |
Camera-perception software for automotive ADAS production programs Sourced | StradVision (스트라드비젼) | L5 | 🇰🇷 | Product revenue not established | 68.1 | 68 | 75 | 62 | |
AI image generation for artists Sourced | Midjourney | L5 | 🌐 | Active-user count not verified; see recorded sources | 67.9 | 85 | 72.2 | 51 | |
Customer agents across chat, voice and messaging AppSourced | Sierra | L5 | 🌐 | $200M+ ARR · 2026-08-31 | 67.2 | 85 | 79.5 | 45 | |
Build software with AI Agent harnessSourced | Replit | L4 | 🌐 | AI Agent active users not disclosed | 67.1 | 82 | 72.2 | 51 | |
Inference platform for open and fine-tuned models Sourced | Fireworks AI | L3 | 🌐 | $1B+ annualized revenue · 2026-07-16 | 66.9 | 74 | 90 | 45 | |
AI video generation Sourced | L5 | 🌐 | Not disclosed | 66.6 | 87 | 50 | 68 | Runway Kling (可灵) | |
Enterprise foundation models Sourced | AWS | L3 | 🌐 | Product revenue not established | 66.3 | 89 | 42 | 78 | |
ByteDance's enterprise model family on Volcano Engine Ark ModelSourced | ByteDance (字节跳动) | L2 | 🇨🇳 | Product revenue not established | 66.2 | 83 | 50 | 70 | |
Video generation model with audio, references and editing ModelSourced | ByteDance (字节跳动) | L2 | 🇨🇳 | Not disclosed | 66.1 | 85 | 50 | 68 | - |
Managed inference on customer-owned or hosted compute PlatformSourced | Inferact | L3 | 🌐 | Product revenue not established | 65.8 | 78 | 60 | 61 | - |
AI-powered cancer detection Sourced | Lunit (루닛) | L5 | 🇰🇷 | ₩22.7B annual revenue · 2025-12-31 | 65.7 | 79 | 59.7 | 60 | |
AI-powered enterprise knowledge agent Sourced | Glean | L5 | 🌐 | $300M+ ARR · 2026-05-28 | 65.6 | 80 | 82.2 | 43 | |
AI video generation and creative editing suite Partly sourced | Runway | L5 | 🌐 | Active-user count not verified; see recorded sources | 65.6 | 84 | 60 | 56 | |
Cerebras wafer-scale AI system, from WSE-3 chip to Condor Galaxy clusters Sourced | Cerebras | L1 | 🌐 | Product revenue not established | 65.4 | 80 | 50 | 70 | |
Google Cloud's fifth-generation TPU accelerators — v5e and v5p Sourced | Google Cloud | L1 | 🌐 | Product revenue not established | 64.6 | 83 | 50 | 65 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 AWS Trainium |
Mistral AI's general-purpose language model family ModelSourced | Mistral AI | L2 | 🌐 | Product revenue not established | 63.8 | 80 | 50 | 65 | |
FDA-cleared deep-learning MRI reconstruction. Sourced | AIRS Medical (에어스메디컬) | L5 | 🇰🇷 | Product revenue not established | 63.8 | 80 | 50 | 65 | |
Huawei's Ascend 910 AI accelerator series — 910B and 910C Sourced | Huawei (华为) | L1 | 🇨🇳 | Product revenue not established | 63 | 77 | 50 | 65 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
Open-source workflow automation and AI agent platform Sourced | n8n | L4 | 🌐 | $40M+ ARR · 2025-12-01 | 62.8 | 82 | 67 | 45 | |
AI diagnostics for medical imaging Sourced | Vuno (뷰노) | L5 | 🇰🇷 | ≥₩25.7B annual revenue · 2025-12-31 · component only | 62.5 | 73 | 60.8 | 55 | |
An inference accelerator evaluated as a complete serving system Sourced | FuriosaAI (퓨리오사AI) | L1 | 🇰🇷 | Product revenue not established | 62.4 | 81 | 50 | 60 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 |
Kuaishou AI video-generation product Sourced | Kuaishou (快手) | L5 | 🇨🇳 | Active-user count not verified; see recorded sources | 62.2 | 82 | 60 | 49 | |
AI answer engine for cited search and research. AppSourced | Perplexity | L5 | 🌐 | Active-user count not verified; see recorded sources | 62.2 | 82 | 60 | 49 | |
Revenue intelligence over recorded sales conversations Sourced | Gong | L5 | 🌐 | $500M+ ARR · 2026-05-12 | 62.1 | 70 | 85.5 | 40 | |
AI SRE agents for incidents and production operations PlatformSourced | Resolve AI | L4 | 🌐 | Product revenue not established | 62 | 78 | 60 | 51 | - |
Moonshot AI's assistant and open-weight K-series frontier models ModelSourced | Moonshot AI | L2 | 🇨🇳 | Not disclosed | 61.4 | 84 | 60 | 46 | |
Training data, human feedback and model evaluation PlatformSourced | Scale AI | L3 | 🌐 | Product revenue not established | 61.3 | 82 | 42 | 67 | - |
AI virtual metrology for semiconductor fabs Sourced | Gauss Labs (가우스랩스) | L3 | 🇰🇷 | Product revenue not established | 61.2 | 74 | 50 | 62 | |
Tencent foundation model family ModelSourced | Tencent (腾讯) | L2 | 🇨🇳 | Product revenue not established | 61.1 | 77 | 45.6 | 65 | |
AI rare-disease whole-exome interpretation. Sourced | 3billion (쓰리빌리언) | L5 | 🇰🇷 | Product revenue not established | 60.9 | 73 | 50 | 62 | |
AI creation and editing inside Canva Partly sourced | Canva | L5 | 🌐 | Not disclosed | 60.7 | 83 | 60 | 45 | |
Suno's AI music generation product Partly sourced | Suno | L5 | 🌐 | Active-user count not verified; see recorded sources | 60.5 | 82 | 60 | 45 | |
Rebellions ATOM inference systems from single server to rack-scale POD Sourced | Rebellions (리벨리온) | L1 | 🇰🇷 | Product revenue not established | 60.3 | 63 | 60 | 58 | |
AI brand protection against counterfeits and IP infringement Sourced | MarqVision (마크비전) | L5 | 🇰🇷 | $22M+ ARR · 2025-06-30 | 59.8 | 77 | 61.8 | 45 | |
Z.ai (Zhipu) GLM model family for agentic engineering ModelSourced | Zhipu AI (智谱AI) | L2 | 🇨🇳 | Not disclosed | 59.6 | 82 | 60 | 43 | |
Meta open-weight models for existing and self-hosted deployments Sourced | Meta | L2 | 🌐 | Current active deployment count not disclosed | 59.5 | 71 | 76 | 39 | |
Korea's sovereign AI foundation model Sourced | Naver (네이버) | L2 | 🇰🇷 | Product revenue not established | 59.4 | 70 | 50 | 60 | |
Photo-based math problem solving and tutoring Sourced | Mathpresso (매스프레소) | L5 | 🇰🇷 | 8M monthly users claimed · measurement date undisclosed | 58.8 | 65 | 94.9 | 33 | |
Google's Apache-licensed open model family for local and edge AI Sourced | L2 | 🌐 | 400M+ family downloads (Google, Apr 2026; not active users) | 58.2 | 82 | 60 | 40 | ||
De-facto orchestration framework for LLM applications Sourced | LangChain | L4 | 🌐 | Not disclosed | 57.9 | 77 | 72.2 | 35 | |
Groq's LPU inference hardware, from chip to node and rack Sourced | Groq | L1 | 🌐 | Product revenue not established | 57.8 | 73 | 45.6 | 58 | |
Lovable's prompt-to-app builder Partly sourced | Lovable | L4 | 🌐 | Not disclosed | 57.7 | 80 | 60 | 40 | |
Japanese language-model family for API and private deployment ModelSourced | Preferred Networks | L2 | 🌐 | Product revenue not established | 57.7 | 74 | 50 | 52 | - |
Open-source LLM application development platform Partly sourced | Dify | L4 | 🌐 | Not disclosed | 57.5 | 77 | 55 | 45 | |
Local and hosted runtime for open models PlatformSourced | Ollama | L3 | 🌐 | Not disclosed | 57.4 | 82 | 42 | 55 | - |
Dark-web threat intelligence and security analytics Sourced | S2W | L3 | 🇰🇷 | Product revenue not established | 57.2 | 77 | 42 | 58 | |
Qwen model-building and application platform Sourced | Alibaba (阿里巴巴) | L2 | 🇨🇳 | Product revenue not established | 57 | 80 | 42 | 55 | |
Agent-centered enterprise model platform from Baidu AI Cloud Sourced | Baidu (百度) | L2 | 🇨🇳 | Product revenue not established | 57 | 80 | 42 | 55 | |
MiniMax's M-series reasoning and agentic model family ModelSourced | MiniMax | L2 | 🇨🇳 | Not disclosed | 57 | 80 | 55 | 42 | |
3 kW FuriosaAI inference appliance for agentic systems Sourced | FuriosaAI (퓨리오사AI) | L1 | 🇰🇷 | Product revenue not established | 56.9 | 67 | 50 | 55 | |
Enterprise LLM assistant over internal documents Sourced | Allganize (올거나이즈) | L5 | 🇰🇷 | Product revenue not established | 56.8 | 71 | 60 | 43 | |
Operate industrial AI across constrained enterprise environments PlatformSourced | MakinaRocks (마키나락스) | L3 | 🇰🇷 | Product revenue not established | 56.7 | 76 | 50 | 48 | MLflow Backend.AI (백엔드.AI) |
Enterprise language models with API and deployment choices ModelSourced | Upstage (업스테이지) | L2 | 🇰🇷 | Product revenue not established | 56.5 | 80 | 50 | 45 | Claude Qwen (通义千问) |
Synthetic avatar video and translation Sourced | HeyGen | L5 | 🌐 | $200M+ ARR · 2026-06-25 | 56.3 | 66 | 79.5 | 34 | |
Lablup's GPU cluster orchestration platform, self-hosted or managed Sourced | Lablup (래블업) | L3 | 🇰🇷 | ≈₩6.6B annual revenue · 2025-12-31 | 56 | 80 | 48.9 | 45 | |
CXL controller and switch IP for AI infrastructure. Partly sourced | Panmnesia (파네시아) | L3 | 🇰🇷 | Product revenue not established | 56 | 76 | 42 | 55 | |
Trace and evaluate agents in actual service workflows PlatformPartly sourced | AIWORKX (에이아이웍스) | L3 | 🇰🇷 | Product revenue not established | 55.7 | 77 | 50 | 45 | LangSmith Langfuse |
Korean foundation-model program with internal LG deployment and open-weight releases Sourced | LG AI Research (LG AI연구원) | L2 | 🇰🇷 | Not disclosed | 55.7 | 71 | 42 | 58 | |
Synthesia's AI avatar video platform for enterprise communication Sourced | Synthesia | L5 | 🌐 | ≈$140M ARR · 2026-02-10 | 55.7 | 66 | 77.2 | 34 | |
Wafer-scale inference cloud for low-latency LLM serving Sourced | Cerebras | L3 | 🌐 | Product revenue not established | 55.6 | 80 | 45.6 | 47 | |
Developer inference cloud powered by Groq LPU silicon Sourced | Groq | L3 | 🌐 | Product revenue not established | 55.6 | 80 | 47.8 | 45 | |
AI wildfire detection on utility camera networks Sourced | Alchera (알체라) | L3 | 🇰🇷 | Product revenue not established | 55.5 | 74 | 42 | 55 | |
3D machine vision and robot-planning suite for industrial automation Sourced | Mech-Mind Robotics (梅卡曼德机器人) | L5 | 🇨🇳 | Product revenue not established | 55.5 | 74 | 42 | 55 | |
Autonomous drone AI for wind, energy, and infrastructure inspection. Sourced | Nearthlab (니어스랩) | L5 | 🇰🇷 | Product revenue not established | 55.5 | 68 | 42 | 60 | |
Robot controllers and fleet software for industrial mobile robots Sourced | SEER Robotics (仙工智能) | L5 | 🇨🇳 | Product revenue not established | 55.5 | 74 | 42 | 55 | |
AI imaging triage for lung, chest and stroke workflows Sourced | Infervision (推想医疗) | L5 | 🇨🇳 | Product revenue not established | 55.2 | 73 | 42 | 55 | |
AI hospital early-warning system for patient deterioration. Sourced | AITRICS (에이아이트릭스) | L5 | 🇰🇷 | Product revenue not established | 55.2 | 73 | 42 | 55 | |
AI-assisted annotation and dataset operations PlatformPartly sourced | AIWORKX (에이아이웍스) | L3 | 🇰🇷 | Product revenue not established | 55 | 74 | 50 | 45 | CVAT / Label Studio Superb Platform |
Infrastructure-based autonomous-vehicle control system. Sourced | Seoul Robotics (서울로보틱스) | L5 | 🇰🇷 | Product revenue not established | 54.9 | 68 | 42 | 58 | |
Ambient clinical documentation inside the EHR Partly sourced | Abridge | L5 | 🌐 | Product revenue not established | 54.7 | 82 | 50 | 40 | |
Expressive AI voice and synthetic-actor studio. Sourced | Neosapience (네오사피엔스) | L5 | 🇰🇷 | Named weekly production use · KLPNA, Aug 2026 | 54.7 | 70 | 60 | 39 | |
Licensable Korean NPU IP for edge-AI chips Sourced | OPENEDGES Technology (오픈엣지테크놀로지) | L1 | 🇰🇷 | Product revenue not established | 54.6 | 67 | 42 | 58 | |
AI research agent and cited search engine. Sourced | Liner (라이너) | L5 | 🇰🇷 | Named enterprise customers · Aug 2026; MAU undisclosed | 54.6 | 68 | 60 | 40 | |
Digital-twin simulation for autonomous driving Sourced | MORAI (모라이) | L5 | 🇰🇷 | Product revenue not established | 54.5 | 74 | 42 | 52 | |
Vision AI data management Sourced | Superb AI (슈퍼브에이아이) | L3 | 🇰🇷 | Product revenue not established | 54.5 | 77 | 42 | 50 | |
Desktop workspace and runtime for open models AppSourced | LM Studio | L3 | 🌐 | Not disclosed | 54.4 | 80 | 42 | 48 | - |
Korean call, meeting and everyday assistant from SK Telecom Sourced | SK Telecom (SK텔레콤) | L5 | 🇰🇷 | 1.10M MAU · est. South Korea app, Sep 2026 | 54.2 | 78 | 63.8 | 32 | chatgpt-for-kakao ChatGPT Gemini |
Unified LLM routing and inference marketplace Sourced | OpenRouter | L3 | 🌐 | Product revenue not established | 54.1 | 79 | 47.8 | 42 | |
3D perception software that turns infrastructure LiDAR into AI insights. Sourced | Seoul Robotics (서울로보틱스) | L5 | 🇰🇷 | Product revenue not established | 53.8 | 74 | 42 | 50 | |
AMD's adaptive SoC/FPGA platform for edge AI inference in industrial, automotive, and aerospace deployments. Sourced | AMD | L1 | 🌐 | Product revenue not established | 53.7 | 67 | 42 | 55 | |
AI detection of cerebral aneurysms on brain MRA Sourced | Deepnoid (딥노이드) | L5 | 🇰🇷 | Product revenue not established | 53.5 | 73 | 42 | 50 | |
Industrial vision models connected to inspection operations PlatformPartly sourced | AIVEX | L3 | 🇰🇷 | Product revenue not established | 53.3 | 80 | 42 | 45 | SAIGE VISION Cognex |
Train and deploy visual defect inspection on production lines PlatformPartly sourced | SAIGE (세이지) | L3 | 🇰🇷 | Product revenue not established | 53.1 | 79 | 42 | 45 | Cognex Superb Platform |
Hardware-agnostic robot navigation stack Sourced | Clobot (클로봇) | L5 | 🇰🇷 | Product revenue not established | 53 | 74 | 42 | 48 | |
Enterprise face-verification engine Sourced | Alchera (알체라) | L3 | 🇰🇷 | Product revenue not established | 52.8 | 78 | 42 | 45 | |
AI document and drawing workflows for AEC firms PlatformSourced | Nomic AI | L3 | 🌐 | Product revenue not established | 52.5 | 75 | 42 | 46 | - |
Enterprise generative-AI platform from LG CNS Sourced | LG CNS (LG씨엔에스) | L4 | 🇰🇷 | Product revenue not established | 52.4 | 76 | 42 | 45 | |
LLM app and agent builder platform Sourced | Wanted Lab (원티드랩) | L4 | 🇰🇷 | Product revenue not established | 52.4 | 76 | 42 | 45 | |
Swarm light-show drone flown in fleets of thousands Sourced | UVify (유비파이) | L5 | 🇰🇷 | Product revenue not established | 52.2 | 65 | 42 | 52 | |
Clinical and research assistant for Tempus reports and records Partly sourced | Tempus | L5 | 🌐 | Product revenue not established | 52.2 | 66 | 50 | 43 | |
Conversational AI for NLU, speech, and personalization. Partly sourced | Skelter Labs (스켈터랩스) | L5 | 🇰🇷 | Product revenue not established | 52.1 | 75 | 42 | 45 | |
AI customer-support agent inside Channel Talk. Sourced | Channel Corp (채널코퍼레이션) | L5 | 🇰🇷 | Product revenue not established | 52.1 | 75 | 42 | 45 | |
AI business messenger for SMBs. Sourced | Channel Corp (채널코퍼레이션) | L5 | 🇰🇷 | Product revenue not established | 52.1 | 75 | 42 | 45 | |
Demand and materials forecasts for ordering decisions PlatformPartly sourced | Impactive AI (임팩티브AI) | L5 | 🇰🇷 | Product revenue not established | 52.1 | 75 | 42 | 45 | Existing planning system Seasonal statistical forecast |
Chat API platform (7B convos/month) Sourced | Sendbird (센드버드) | L3 | 🇰🇷 | Product revenue not established | 52.1 | 75 | 42 | 45 | Fin Zendesk AI twilio |
Omnichannel AI customer service and agent platform Sourced | Udesk (沃丰科技) | L5 | 🇨🇳 | Product revenue not established | 52.1 | 75 | 42 | 45 | |
Korean enterprise and public-sector AI chatbot Sourced | Wisenut (와이즈넛) | L5 | 🇰🇷 | Product revenue not established | 52.1 | 75 | 42 | 45 | |
Edge-AI NPU for industrial vision and robotics. Sourced | Mobilint (모빌린트) | L1 | 🇰🇷 | Product revenue not established | 52 | 67 | 42 | 50 | |
AI assistant across Meta apps and the web Sourced | Meta | L5 | 🌐 | ~1B MAU across Meta apps (Meta-reported, bundled cross-app figure) | 52 | 78 | 25 | 72 | ChatGPT Gemini Microsoft Copilot |
Decentralized GPU cloud for AI training and inference. Partly sourced | Common Computer (커먼컴퓨터) | L3 | 🇰🇷 | Product revenue not established | 51.9 | 74 | 42 | 45 | |
Browser-based AI dental restoration CAD with editable crown designs Sourced | Imagoworks (이마고웍스) | L5 | 🇰🇷 | Product revenue not established | 51.9 | 74 | 42 | 45 | |
AI for renewable-energy site selection and PPA structuring. Sourced | Energy X (에너지엑스) | L3 | 🇰🇷 | Product revenue not established | 51.9 | 74 | 42 | 45 | |
Self-hostable Friendli Engine for inference in your own environment. Sourced | FriendliAI | L3 | 🇰🇷 | Product revenue not established | 51.9 | 74 | 42 | 45 | |
Managed APIs and dedicated capacity for open models PlatformSourced | FriendliAI | L3 | 🇰🇷 | Product revenue not established | 51.9 | 74 | 42 | 45 | Together / Fireworks vLLM on rented GPUs |
AMD's open-source GPU computing platform — the software layer enabling AI and HPC workloads on Instinct accelerators. Sourced | AMD | L3 | 🌐 | Product revenue not established | 51.9 | 74 | 42 | 45 | |
Tesla's vision-only autonomous driving stack Partly sourced | Tesla | L5 | 🌐 | Active-user count not verified; see recorded sources | 51.9 | 68 | 25 | 82 | |
VESSL's MLOps platform for training, serving, and pipelines. Sourced | VESSL AI (베슬에이아이) | L3 | 🇰🇷 | Product revenue not established | 51.9 | 74 | 42 | 45 | weights-biases MLflow neptune |
AI-native risk, lending and service stack for digital banking Sourced | WeBank (微众银行) | L5 | 🇨🇳 | Product revenue not established | 51.9 | 74 | 42 | 45 | |
Connect enterprise data to supervised agent workflows PlatformPartly sourced | DeepAuto (딥오토) | L4 | 🇰🇷 | Product revenue not established | 51.5 | 74 | 42 | 44 | Gemini Claude |
Autonomous indoor delivery robot. Sourced | ROBOTIS (로보티즈) | L5 | 🇰🇷 | Product revenue not established | 51.5 | 65 | 42 | 50 | |
Wearable walking-assist robot generating real-world movement data. Sourced | WIRobotics (위로보틱스) | L5 | 🇰🇷 | Not disclosed | 51.5 | 65 | 42 | 50 | |
Korea's consumer AI platform aggregating frontier models Sourced | Wrtn (뤼튼) | L5 | 🇰🇷 | 380K MAU · est. South Korea app, Sep 2026 | 51.5 | 78 | 54.6 | 32 | ChatGPT Perplexity |
Amazon's multimodal foundation-model family in Bedrock Partly sourced | AWS | L2 | 🌐 | Product revenue not established | 51.4 | 80 | 25 | 68 | |
Model compression and hardware-aware inference optimization PlatformSourced | SqueezeBits (스퀴즈비츠) | L3 | 🇰🇷 | Product revenue not established | 51.4 | 72 | 42 | 45 | vLLM / SGLang Friendli Inference Cloud |
Find and repair gaps in AI training data PlatformPartly sourced | Pebblous (페블러스) | L3 | 🇰🇷 | Product revenue not established | 51.1 | 72 | 42 | 44 | Cleanlab Superb Platform |
Video understanding API Sourced | Twelve Labs (트웰브랩스) | L3 | 🇰🇷 | Product revenue not established | 51 | 59 | 50 | 45 | |
Baidu's ERNIE model and assistant family ModelSourced | Baidu (百度) | L2 | 🇨🇳 | Product revenue not established | 50.9 | 80 | 25 | 66 | |
Meta’s Muse model family, separate from Llama ModelSourced | Meta | L2 | 🌐 | Not disclosed | 50.7 | 80 | 25 | 65 | - |
Conversational AI shopping agent for e-commerce. Partly sourced | Waddle (와들) | L5 | 🇰🇷 | Product revenue not established | 50.5 | 73 | 42 | 42 | |
IONIQ 5 vehicle-platform integration for Waymo One autonomous ride-hailing. Sourced | Hyundai (현대자동차) | L5 | 🇰🇷 | Product revenue not established | 50.5 | 68 | 42 | 45 | |
AI fashion image auto-tagging. Partly sourced | OMNIOUS.AI (옴니어스) | L5 | 🇰🇷 | Product revenue not established | 50.5 | 73 | 42 | 42 | |
Enterprise QA on machine reading comprehension. Partly sourced | 42Maru (포티투마루) | L5 | 🇰🇷 | Product revenue not established | 50.4 | 71 | 42 | 43 | |
AI CRM for D2C merchants on Cafe24 / Naver SmartStore. Sourced | Datarize (데이터라이즈) | L5 | 🇰🇷 | Product revenue not established | 50.4 | 76 | 42 | 40 | |
Duolingo's AI tier with roleplay, explanations and video call Sourced | Duolingo | L5 | 🌐 | Not disclosed | 50.4 | 73 | 50 | 35 | |
Korean enterprise generative-AI and RAG platform. Partly sourced | Sionic AI (사이오닉에이아이) | L4 | 🇰🇷 | Product revenue not established | 50.4 | 71 | 42 | 43 | |
AI development and operations delivered with implementation support PlatformSourced | ABEJA | L3 | 🌐 | Product revenue not established | 50.2 | 70 | 42 | 43 | - |
Industrial AI PC with 80 TOPS NPU. Partly sourced | Mobilint (모빌린트) | L1 | 🇰🇷 | Product revenue not established | 50.2 | 67 | 42 | 45 | |
FDA-cleared AI OB/GYN ultrasound — identifies 39 standard views, labels 47 structures, reduces keystrokes 94%. Partly sourced | Samsung Medison | L5 | 🇰🇷 | Product revenue not established | 50.2 | 78 | 25 | 65 | GE HealthCare Philips Siemens Healthineers |
Face recognition and computer-vision APIs and SDKs PlatformSourced | Megvii (Face++) (旷视科技) | L3 | 🇨🇳 | Product revenue not established | 50.1 | 68 | 42 | 44 | - |
AI detection across Proofpoint's threat protection platform Partly sourced | Proofpoint | L3 | 🌐 | Product revenue not established | 50 | 77 | 25 | 65 | |
Identity verification and fraud platform Partly sourced | Socure | L5 | 🌐 | Product revenue not established | 49.4 | 78 | 25 | 62 | |
Tesla's humanoid robot Partly sourced | Tesla | L5 | 🌐 | Product revenue not established | 49.2 | 55 | 30 | 72 | |
Open-source desktop assistant for local or remote models AppSourced | Jan | L3 | 🌐 | Not disclosed | 49.1 | 74 | 42 | 38 | - |
Amazon Bedrock foundation-model family Partly sourced | AWS | L2 | 🌐 | Product revenue not established | 48.8 | 80 | 25 | 58 | |
Cohere's enterprise model family for RAG and tool use Partly sourced | Cohere | L2 | 🌐 | Product revenue not established | 48.8 | 80 | 25 | 58 | |
Hosted model APIs for language, video, speech and music PlatformSourced | MiniMax | L3 | 🇨🇳 | Product revenue not established | 48.8 | 80 | 25 | 58 | |
Multimodal AI video-interview assessment Sourced | Genesis Lab (제네시스랩) | L5 | 🇰🇷 | Product revenue not established | 48.8 | 69 | 42 | 40 | |
AI software and programming stack for Doosan collaborative robots Partly sourced | Doosan Robotics (두산로보틱스) | L5 | 🇰🇷 | Product revenue not established | 48.6 | 74 | 25 | 62 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
Korean deep-learning text-to-speech suite Sourced | Selvas AI (셀바스AI) | L5 | 🇰🇷 | Product revenue not established | 48.6 | 70 | 42 | 39 | |
BGM subscription powered by Pozalabs AI. Partly sourced | Pozalabs (포자랩스) | L5 | 🇰🇷 | Not disclosed | 48.6 | 70 | 42 | 39 | |
Cambricon's datacentre AI accelerators and Neuware toolchain Partly sourced | Cambricon | L1 | 🇨🇳 | Product revenue not established | 48.5 | 76 | 25 | 60 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
Classroom Q&A platform with AI clustering and teaching assistance Sourced | CLASSUM (클라썸) | L5 | 🇰🇷 | Product revenue not established | 48.5 | 65 | 50 | 35 | |
Figure AI's general-purpose humanoid robot line Partly sourced | Figure AI | L5 | 🌐 | Product revenue not established | 48.5 | 65 | 25 | 70 | |
Academic research AI for paper-based research. Sourced | Liner (라이너) | L5 | 🇰🇷 | Yonsei access agreement · Mar 2026; module usage undisclosed | 48.5 | 68 | 42 | 40 | |
Vision-AI food scanning and dietary analytics Sourced | Nuvilab (누비랩) | L5 | 🇰🇷 | Product revenue not established | 48.5 | 68 | 42 | 40 | |
MFDS-approved AI digital therapeutic for insomnia. Sourced | WELT (웰트) | L5 | 🇰🇷 | Product revenue not established | 48.5 | 68 | 42 | 40 | |
Live multilingual captioning and translation for events. Partly sourced | XL8 (엑스엘에이트) | L5 | 🇰🇷 | Product revenue not established | 48.3 | 69 | 42 | 39 | |
Media-localization machine translation for film and TV. Partly sourced | XL8 (엑스엘에이트) | L5 | 🇰🇷 | Product revenue not established | 48.3 | 69 | 42 | 39 | |
Multilingual lip-sync dubbing for video. Sourced | ESTsoft (이스트소프트) | L5 | 🇰🇷 | Product revenue not established | 48.3 | 69 | 42 | 39 | |
Korean/Japanese speech-to-text API Sourced | Return Zero (리턴제로) | L3 | 🇰🇷 | Product revenue not established | 48.3 | 69 | 42 | 39 | |
Korean-optimized enterprise speech-to-text engine Sourced | Selvas AI (셀바스AI) | L5 | 🇰🇷 | Product revenue not established | 48.3 | 69 | 42 | 39 | |
Synthetic AI humans for kiosks, video and virtual presenters Sourced | DeepBrain AI (딥브레인AI) | L5 | 🇰🇷 | Product revenue not established | 48.2 | 66 | 50 | 34 | |
Sequence-based AI drug discovery platform Sourced | Deargen (디어젠) | L5 | 🇰🇷 | Product revenue not established | 48 | 55 | 42 | 48 | Standigm ASK Chemistry42 HyperLab (하이퍼랩) |
De novo antibody and protein design platform. Partly sourced | Galux (갤럭스) | L5 | 🇰🇷 | Product revenue not established | 48 | 55 | 42 | 48 | |
Transcriptomics-based AI drug repositioning Sourced | Oncocross (온코크로스) | L5 | 🇰🇷 | Product revenue not established | 48 | 55 | 42 | 48 | |
AI drug discovery platform. Partly sourced | Standigm (스탠다임) | L5 | 🇰🇷 | Product revenue not established | 48 | 55 | 42 | 48 | |
Enterprise AI services PlatformPartly sourced | Azure | L3 | 🌐 | Product revenue not established | 47.9 | 80 | 25 | 55 | |
Baichuan's model API and enterprise agent platform Partly sourced | Baichuan AI | L2 | 🇨🇳 | Product revenue not established | 47.9 | 80 | 25 | 55 | |
Mistral AI production platform for building and operating AI systems Partly sourced | Mistral AI | L2 | 🌐 | Product revenue not established | 47.9 | 80 | 25 | 55 | |
Managed AI training capacity on NVIDIA GPU clusters, rented through major clouds Sourced | NVIDIA | L3 | 🌐 | Product revenue not established | 47.9 | 80 | 25 | 55 | Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p AWS Trainium |
Developer access for StepFun models Partly sourced | StepFun | L2 | 🇨🇳 | Product revenue not established | 47.9 | 80 | 25 | 55 | |
Tencent Cloud's end-to-end machine-learning platform Partly sourced | Tencent (腾讯) | L2 | 🇨🇳 | Product revenue not established | 47.9 | 80 | 25 | 55 | |
Post-training infrastructure behind a Python API PlatformSourced | Thinking Machines Lab | L3 | 🌐 | Product revenue not established | 47.9 | 80 | 25 | 55 | - |
ByteDance cloud model and agent platform Partly sourced | ByteDance (字节跳动) | L2 | 🇨🇳 | Product revenue not established | 47.9 | 80 | 25 | 55 | |
Z.ai API and model platform Partly sourced | Zhipu AI (智谱AI) | L2 | 🇨🇳 | Product revenue not established | 47.9 | 80 | 25 | 55 | |
Market intelligence search across filings, transcripts and expert calls Partly sourced | AlphaSense | L5 | 🌐 | Product revenue not established | 47.7 | 69 | 25 | 63 | |
LLM safety and de-identification API. Partly sourced | TUNiB (튜닙) | L3 | 🇰🇷 | Product revenue not established | 47.7 | 68 | 42 | 38 | |
Spatial AI digital twin platform. Partly sourced | 3i Inc. (쓰리아이) | L5 | 🇰🇷 | Product revenue not established | 47.6 | 66 | 42 | 39 | |
Financial crime and transaction fraud prevention platform Partly sourced | Feedzai | L5 | 🌐 | Product revenue not established | 47.5 | 78 | 25 | 55 | |
Fraud, compliance and risk platform with device intelligence Partly sourced | Sardine | L5 | 🌐 | Product revenue not established | 47.5 | 78 | 25 | 55 | |
AI for financial crime Partly sourced | Symphony AyasdiAI | L5 | 🌐 | Product revenue not established | 47.5 | 78 | 25 | 55 | |
Behavioural AI against email compromise and account takeover Partly sourced | Abnormal Security | L3 | 🌐 | Product revenue not established | 47.3 | 77 | 25 | 55 | |
Continuous blood-pressure monitoring ring. Sourced | Sky Labs (스카이랩스) | L5 | 🇰🇷 | Product revenue not established | 47.3 | 73 | 25 | 58 | |
Embedding, reranking and retrieval tooling for enterprise RAG Partly sourced | Cohere | L2 | 🌐 | Product revenue not established | 47.3 | 77 | 25 | 55 | |
Self-learning network anomaly detection and response Partly sourced | Darktrace | L3 | 🌐 | Product revenue not established | 47.3 | 77 | 25 | 55 | |
Unified data and AI platform Sourced | Databricks | L3 | 🌐 | Product revenue not established | 47.3 | 77 | 25 | 55 | |
End-to-end data science and AI platform for enterprises Partly sourced | Dataiku | L3 | 🌐 | Product revenue not established | 47.3 | 77 | 25 | 55 | |
Enterprise AI platform for AutoML, apps and observability Partly sourced | DataRobot | L3 | 🌐 | Product revenue not established | 47.3 | 77 | 25 | 55 | |
Programmable smart actuators used in robotics research worldwide. Sourced | ROBOTIS (로보티즈) | L5 | 🇰🇷 | Product revenue not established | 47.3 | 65 | 25 | 65 | |
AI across Elastic's search, observability and security stack Partly sourced | Elastic | L3 | 🌐 | Product revenue not established | 47.3 | 77 | 25 | 55 | |
AutoML, document AI and open LLMs on one platform Partly sourced | H2O.ai | L3 | 🌐 | Product revenue not established | 47.3 | 77 | 25 | 55 | |
Alibaba's open model hub and Model-as-a-Service community Sourced | Alibaba (阿里巴巴) | L3 | 🇨🇳 | Not disclosed | 47.3 | 77 | 25 | 55 | |
Developer security scanning with AI code analysis Partly sourced | Snyk | L3 | 🌐 | Product revenue not established | 47.3 | 77 | 25 | 55 | |
The agile mobile robot Partly sourced | Boston Dynamics | L5 | 🌐 | Product revenue not established | 47.3 | 65 | 25 | 65 | Unitree Go2 Datacenter GPU family Instinct GPU MI300X/MI350/MI400 |
AI-driven network detection and response Partly sourced | Vectra AI | L3 | 🌐 | Product revenue not established | 47.3 | 77 | 25 | 55 | |
Weaviate's open-source vector database Partly sourced | Weaviate | L3 | 🌐 | Product revenue not established | 47.3 | 77 | 25 | 55 | |
Cloud security graph with AI posture management Partly sourced | Wiz | L3 | 🌐 | Product revenue not established | 47.3 | 77 | 25 | 55 | |
Software-defined vehicle platform for Hyundai Motor Group. Sourced | 42dot (포티투닷) | L5 | 🇰🇷 | Product revenue not established | 47.2 | 68 | 25 | 62 | |
Open-source coding harness with model-provider choice Agent harnessSourced | Anomaly | L4 | 🌐 | Not disclosed | 47.2 | 78 | 42 | 32 | - |
PCIe 6.4 / CXL 3.2 fabric switch for AI servers. Partly sourced | Panmnesia (파네시아) | L3 | 🇰🇷 | Product revenue not established | 47.1 | 76 | 25 | 55 | |
AI webtoon production studio. Partly sourced | LionRocket (라이온로켓) | L5 | 🇰🇷 | Product revenue not established | 46.9 | 70 | 42 | 35 | |
Baidu autonomous-driving and intelligent-vehicle platform Partly sourced | Baidu (百度) | L2 | 🇨🇳 | Not disclosed | 46.7 | 68 | 25 | 60 | |
Hanwha Aerospace multipurpose unmanned ground vehicle. Partly sourced | Hanwha (한화) | L5 | 🇰🇷 | Product revenue not established | 46.7 | 68 | 25 | 60 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
Huawei's Atlas AI compute systems for datacentre and edge Partly sourced | Huawei (华为) | L1 | 🇨🇳 | Product revenue not established | 46.7 | 74 | 25 | 55 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
Horizon Robotics' assisted-driving stack on Journey silicon Partly sourced | Horizon Robotics | L5 | 🇨🇳 | Product revenue not established | 46.7 | 68 | 25 | 60 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
Huawei's open-source deep-learning framework for Ascend Partly sourced | Huawei (华为) | L3 | 🇨🇳 | Product revenue not established | 46.7 | 74 | 25 | 55 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
L4 autonomous-driving software in paid commercial service Sourced | RideFlux (라이드플럭스) | L5 | 🇰🇷 | Product revenue not established | 46.7 | 68 | 25 | 60 | |
Radiology triage platform across clinical specialties Partly sourced | Aidoc | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | |
Retinal image analysis for chronic-disease screening Sourced | Airdoc (鹰瞳科技) | L5 | 🇨🇳 | Product revenue not established | 46.5 | 73 | 25 | 55 | |
AI 12-lead ECG screening for hidden cardiac disease Sourced | Medical AI (메디컬에이아이) | L5 | 🇰🇷 | Product revenue not established | 46.5 | 73 | 25 | 55 | |
FDA Breakthrough–designated brain CT AI for hemorrhage and hematoma — Medicare NTAP reimbursed. Partly sourced | Harrison.ai | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | |
12 FDA-cleared findings from a single chest X-ray — pneumothorax, effusion, vertebral fractures, and more. Partly sourced | Harrison.ai | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | |
First FDA-cleared continuous AI sepsis monitoring — 18% lower mortality when clinicians act on alerts. Partly sourced | Bayesian Health | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | Prenosis Epic Deterioration Index Dascena |
10 FDA clearances for stroke imaging triage — mandated across all NHS England and Wales stroke centers. Partly sourced | Brainomix | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | Rapid Viz.ai One |
AI monitoring of pulmonary fibrosis progression from CT — co-primary endpoint in Boehringer Ingelheim Phase 3 trial. Partly sourced | Brainomix | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | SubtleMR Thirona IMBIO |
AI vessel suppression for chest CT improving nodule visibility; opportunistic CAC scoring from routine ungated CTs. Partly sourced | Riverain Technologies | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | Virtual Nodule Clinic Aidoc SubtleMR |
AI bone suppression and line/tube confirmation for chest X-ray — improves subtle lung lesion detection. Partly sourced | Riverain Technologies | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | Gleamer Aidoc harrison-ai-annalise-cxr |
FDA-cleared digital pathology AI platform for primary cancer diagnosis and pharma drug development. Partly sourced | Proscia | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | |
AI cardiovascular disease risk prediction from a single retinal photograph — MFDS cleared, 8-country international deployment. Partly sourced | Mediwhale | L5 | 🇰🇷 | Product revenue not established | 46.5 | 73 | 25 | 55 | EyeArt LumineticsCore sky-labs-cardi-o |
Three first-of-kind FDA clearances detecting fatal cardiac conditions invisible to human readers on standard ECG. Partly sourced | Anumana | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | Eko ELEFT AliveCor Cardiologs |
FDA-cleared AI stethoscope algorithm detecting low ejection fraction — CMS reimbursed at $128.90/use. Partly sourced | Eko Health | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | ECG-AI Suite AliveCor Cardiologs |
Autonomous diabetic retinopathy screening cleared for two camera manufacturers — Norway's national AI eye screening system. Partly sourced | Eyenuk | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | LumineticsCore Google Health Notal Vision |
Non-invasive coronary physiology from CT — replaces diagnostic catheterization for stable chest pain. Partly sourced | HeartFlow | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | HeartFlow Plaque Analysis Siemens Healthineers cFFR Canon Medical |
AI coronary plaque vulnerability analysis from CTA — identifies rupture-prone plaques; nationally covered by Aetna and Cigna. Partly sourced | HeartFlow | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | HeartFlow FFRCT Analysis Siemens Healthineers cFFR Cleerly |
FDA-cleared AI for stroke imaging triage. Sourced | Heuron (휴런) | L5 | 🇰🇷 | Product revenue not established | 46.5 | 73 | 25 | 55 | |
The first autonomous FDA-cleared AI diagnostic in any medical specialty — diabetic retinopathy without physician review. Partly sourced | Digital Diagnostics | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | |
FDA-cleared real-time coronary angiography AI — vessel segmentation and lesion quantification in 1–2 seconds. Partly sourced | Medipixel | L5 | 🇰🇷 | Product revenue not established | 46.5 | 73 | 25 | 55 | HeartFlow FFRCT Analysis Siemens Healthineers Philips |
AI brain MRI for Alzheimer's diagnostics. Sourced | Neurophet (뉴로핏) | L5 | 🇰🇷 | Product revenue not established | 46.5 | 73 | 25 | 55 | |
AI for digital pathology slide review Partly sourced | Paige | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | |
Korean humanoid and manipulation robotics platform Partly sourced | Rainbow Robotics (레인보우로보틱스) | L5 | 🇰🇷 | Product revenue not established | 46.5 | 65 | 25 | 62 | |
RapidAI's imaging platform for stroke and vascular triage Partly sourced | RapidAI | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | |
MFDS and FDA-cleared AI echocardiography analysis — 70%+ reporting time reduction; ~3,000 institutions worldwide. Partly sourced | Ontact Health | L5 | 🇰🇷 | Product revenue not established | 46.5 | 73 | 25 | 55 | Eko ELEFT ECG-AI Suite Samsung Z20 |
FDA-cleared AI reduces MRI scan time by up to 60% while maintaining diagnostic quality. Partly sourced | Subtle Medical | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | GE HealthCare Siemens Healthineers MR Philips MR |
FDA-cleared AI enabling 4x faster PET scans at lower radiotracer dose — scanner-agnostic across major OEMs. Partly sourced | Subtle Medical | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | GE HealthCare Siemens Healthineers PET Philips PET |
AI for precision medicine Partly sourced | Tempus AI | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | |
AI layer across United Imaging's scanner and radiotherapy fleet Sourced | United Imaging (联影医疗) | L5 | 🇨🇳 | Product revenue not established | 46.5 | 73 | 25 | 55 | |
World's first FDA-cleared radiomics AI for lung cancer risk prediction — CPT-reimbursed at $650/procedure. Partly sourced | Optellum | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | Aidoc ClearRead CT Rad AI Continuity |
Care-coordination platform for time-critical imaging findings Partly sourced | Viz.ai | L5 | 🌐 | Product revenue not established | 46.5 | 73 | 25 | 55 | |
Level 4 autonomous-driving stack for shuttles, robotaxis, and AV trucks. Sourced | Autonomous a2z (오토노머스에이투지) | L5 | 🇰🇷 | Product revenue not established | 46.2 | 68 | 25 | 58 | |
Turn captured objects into editable 3D assets AppPartly sourced | RECON Labs (리콘랩스) | L5 | 🇰🇷 | Not disclosed | 46 | 68 | 42 | 34 | Meshy Photogrammetry workflow |
Hanwha Robotics collaborative robot line. Partly sourced | Hanwha (한화) | L5 | 🇰🇷 | Product revenue not established | 46 | 65 | 25 | 60 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
Big-data credit AI for alt-lending. Sourced | HonestAI (어니스트에이아이) | L5 | 🇰🇷 | Product revenue not established | 46 | 78 | 25 | 50 | |
Korean lawyer marketplace and legal information portal Sourced | Law&Company (로앤컴퍼니) | L5 | 🇰🇷 | Product revenue not established | 46 | 66 | 42 | 35 | |
ML credit scoring for near-prime borrowers. Partly sourced | PeopleFund (피플펀드) | L5 | 🇰🇷 | Product revenue not established | 46 | 78 | 25 | 50 | |
AI legal assistant for Korean lawyers and legal teams Sourced | Law&Company (로앤컴퍼니) | L5 | 🇰🇷 | Product revenue not established | 46 | 66 | 42 | 35 | |
World's most dynamic humanoid Partly sourced | Boston Dynamics | L5 | 🌐 | Product revenue not established | 45.8 | 65 | 25 | 59 | Figure Tesla Optimus Unitree Humanoid |
Build and operate HyperCLOVA X applications on NAVER Cloud Partly sourced | Naver (네이버) | L2 | 🇰🇷 | Product revenue not established | 45.8 | 70 | 25 | 55 | |
AI training-data construction service for LLMs. Sourced | Flitto (플리토) | L3 | 🇰🇷 | Product revenue not established | 45.8 | 77 | 25 | 50 | |
Autonomous mobile robots and orchestration for warehouse fulfilment Partly sourced | Geekplus | L5 | 🇨🇳 | Product revenue not established | 45.8 | 65 | 25 | 59 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
Safe humanoid for human environments Partly sourced | 1X Technologies | L5 | 🌐 | Pilot deployments | 45.8 | 65 | 25 | 59 | |
UBTech's humanoid robot line for service and industrial work Partly sourced | UBTech | L5 | 🇨🇳 | Product revenue not established | 45.8 | 65 | 25 | 59 | |
Affordable quadruped robot Partly sourced | Unitree Robotics (宇树科技) | L5 | 🇨🇳 | Active-user count not verified; see recorded sources | 45.8 | 65 | 25 | 59 | Spot Datacenter GPU family Instinct GPU MI300X/MI350/MI400 |
Unitree's humanoid robot line Partly sourced | Unitree Robotics (宇树科技) | L5 | 🇨🇳 | Product revenue not established | 45.8 | 65 | 25 | 59 | |
FDA-cleared AI for prostate cancer grading. Sourced | Deep Bio (딥바이오) | L5 | 🇰🇷 | Product revenue not established | 45.6 | 73 | 25 | 52 | |
Rebellions server for efficient large-scale AI inference Sourced | Rebellions (리벨리온) | L1 | 🇰🇷 | Product revenue not established | 45.6 | 57 | 32 | 52 | |
Hyundai Motor Group's mobile eccentric droid robotics platform. Partly sourced | Hyundai (현대자동차) | L5 | 🇰🇷 | Product revenue not established | 45.5 | 65 | 25 | 58 | |
Interactive AI avatar interface for kiosks and service environments Sourced | ESTsoft (이스트소프트) | L5 | 🇰🇷 | Product revenue not established | 45.5 | 66 | 42 | 34 | |
Autonomous food-service robot at chain scale Sourced | Bear Robotics (베어로보틱스) | L5 | 🇰🇷 | Product revenue not established | 45.5 | 65 | 25 | 58 | |
Kia's highway driver-assistance feature set. Partly sourced | Kia (기아) | L5 | 🇰🇷 | Not disclosed | 45.4 | 68 | 25 | 55 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
AI meeting and call analysis for teams Sourced | Return Zero (리턴제로) | L5 | 🇰🇷 | Product revenue not established | 45.3 | 67 | 42 | 33 | |
AI-native entertainment and character-chat platform Sourced | Scatter Lab (스캐터랩) | L5 | 🇰🇷 | 1.48M MAU · reported Sep 2026; geography unspecified | 45.3 | 56 | 66.4 | 25 | |
Autonomous-driving and vision data labelling Partly sourced | AIMMO (에이모) | L3 | 🇰🇷 | Product revenue not established | 45.2 | 77 | 25 | 48 | |
Automotive AI compute platform for driver assistance Partly sourced | Black Sesame Technologies | L5 | 🇨🇳 | Product revenue not established | 45.2 | 67 | 25 | 55 | |
4D imaging radar with LiDAR-equivalent perception. Sourced | Deep Fusion AI (딥퓨전에이아이) | L5 | 🇰🇷 | Product revenue not established | 45.2 | 74 | 25 | 50 | |
AMD's integrated NPU for consumer and commercial laptops, enabling Copilot+ PC and on-device AI features. Sourced | AMD | L1 | 🌐 | Consumer and commercial laptop buyers | 45.2 | 67 | 25 | 55 | |
Dexterous robot policies, with deployment still to prove ModelSourced | RLWRLD | L2 | 🇰🇷 | Product revenue not established | 45.1 | 75 | 25 | 49 | NVIDIA Isaac Physical Intelligence Robot Foundation Model |
Korea's first MFDS-approved generative AI radiology report generator — 57 chest X-ray conditions. Partly sourced | Soombit.ai | L5 | 🇰🇷 | Product revenue not established | 45 | 73 | 25 | 50 | coreline-soft lunit-scope-cxr vuno-med-chest-xray |
Coronary artery calcium scoring from CT Partly sourced | Coreline Soft (코어라인소프트) | L5 | 🇰🇷 | Product revenue not established | 45 | 73 | 25 | 50 | |
Quantitative lung texture analysis on CT Partly sourced | Coreline Soft (코어라인소프트) | L5 | 🇰🇷 | Product revenue not established | 45 | 73 | 25 | 50 | |
AI test-prep tutor for TOEIC and TOEFL Sourced | Socra AI | L5 | 🇰🇷 | Not disclosed | 44.8 | 65 | 42 | 33 | |
Korean collaborative robots with AI vision Sourced | Neuromeka (뉴로메카) | L5 | 🇰🇷 | Product revenue not established | 44.7 | 65 | 25 | 55 | |
Open-weight models for agentic software work ModelSourced | Poolside AI | L2 | 🌐 | Not disclosed | 44.7 | 73 | 25 | 49 | - |
Autonomous defense drone built on Nearthlab's industrial-inspection stack. Sourced | Nearthlab (니어스랩) | L5 | 🇰🇷 | Product revenue not established | 44.7 | 65 | 25 | 55 | |
Remote identity verification for Korean finance Sourced | Alchera (알체라) | L5 | 🇰🇷 | Product revenue not established | 44.4 | 78 | 25 | 45 | |
AI KYC / KYB / AML identity verification. Sourced | Argos Identity (아르고스아이덴티티) | L5 | 🇰🇷 | Product revenue not established | 44.4 | 78 | 25 | 45 | |
Medical-grade ring with deep-learning AFib detection. Sourced | Sky Labs (스카이랩스) | L5 | 🇰🇷 | Product revenue not established | 44.4 | 73 | 25 | 48 | |
AI interpretation of satellite imagery PlatformPartly sourced | SI Analytics | L3 | 🇰🇷 | Product revenue not established | 44.4 | 78 | 25 | 45 | GIS + in-house vision models Imagery-provider analytics |
MFDS-approved VR digital therapeutic for stroke. Sourced | Nunaps (누랩스) | L5 | 🇰🇷 | Product revenue not established | 44.4 | 73 | 25 | 48 | |
Retinal-photo AI screening for chronic-disease complications Sourced | ARK (아크) | L5 | 🇰🇷 | Product revenue not established | 44.4 | 73 | 25 | 48 | |
DEEPX's DX-M on-device inference NPU series — DX-M1 and DX-M2 Partly sourced | DEEPX | L1 | 🇰🇷 | Product revenue not established | 44.3 | 67 | 25 | 52 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
Korean-language enterprise search, LLM and analytics for closed networks Sourced | Konan Technology (코난테크놀로지) | L2 | 🇰🇷 | Product revenue not established | 44.2 | 77 | 25 | 45 | |
A pipeline from demonstrations to robot policies PlatformSourced | Config Intelligence (컨피그인텔리전스) | L5 | 🇰🇷 | Product revenue not established | 44.2 | 77 | 25 | 45 | LeRobot + internal data pipeline NVIDIA Isaac |
Enterprise AI agent build-and-deploy platform Sourced | Ant Digital Technologies | L4 | 🇨🇳 | Product revenue not established | 44.1 | 76 | 25 | 45 | |
Deploy and operate agents on AWS Partly sourced | AWS | L4 | 🌐 | Product revenue not established | 44.1 | 76 | 25 | 45 | |
AI video creation and editing tools in CapCut Partly sourced | ByteDance (字节跳动) | L2 | 🇨🇳 | Not disclosed | 44.1 | 76 | 25 | 45 | |
Captions' AI video creation and editing app Partly sourced | Captions | L5 | 🌐 | Not disclosed | 44.1 | 76 | 25 | 45 | |
ByteDance AI agent and bot-building platform Partly sourced | ByteDance (字节跳动) | L2 | 🇨🇳 | Product revenue not established | 44.1 | 76 | 25 | 45 | |
Text-based audio and video editing Partly sourced | Descript | L5 | 🌐 | Not disclosed | 44.1 | 76 | 25 | 45 | |
AI image generation Partly sourced | ByteDance (字节跳动) | L5 | 🇨🇳 | Active-user count not verified; see recorded sources | 44.1 | 76 | 25 | 45 | Midjourney dall-e kolors |
Optimize and deploy models for target devices PlatformSourced | Nota AI (노타AI) | L3 | 🇰🇷 | Product revenue not established | 44.1 | 76 | 25 | 45 | Hardware-vendor SDK / TensorRT Yetter |
Build internal apps, workflows and AI agents on business data PlatformSourced | Retool | L4 | 🌐 | Product revenue not established | 44.1 | 76 | 25 | 45 | |
Adversarial testing for models, retrieval systems and agents PlatformPartly sourced | AIM Intelligence (에임인텔리전스) | L3 | 🇰🇷 | Product revenue not established | 44.1 | 76 | 25 | 45 | Promptfoo / internal red-team suite Lakera |
AI upscaling, denoising and restoration for photo and video Partly sourced | Topaz Labs | L5 | 🌐 | Not disclosed | 44.1 | 76 | 25 | 45 | |
Enterprise brand-conditioned content generation Partly sourced | Typeface | L5 | 🌐 | Product revenue not established | 44.1 | 76 | 25 | 45 | |
Agent builder for enterprise document workflows Sourced | Upstage (업스테이지) | L5 | 🇰🇷 | Product revenue not established | 44.1 | 76 | 25 | 45 | |
Browser video editor with AI subtitles, avatars and generation Partly sourced | VEED | L5 | 🌐 | Not disclosed | 44.1 | 76 | 25 | 45 | |
AI video editor that cuts footage by transcript. Partly sourced | VoyagerX (보이저엑스) | L5 | 🇰🇷 | Not disclosed | 44.1 | 76 | 25 | 45 | |
Agent building and deployment for Korean consumer AI Partly sourced | Wrtn (뤼튼) | L4 | 🇰🇷 | Product revenue not established | 44.1 | 76 | 25 | 45 | |
AI agents and chatbots across Zapier's automation graph Partly sourced | Zapier | L4 | 🌐 | Product revenue not established | 44.1 | 76 | 25 | 45 | |
Phenomics-driven drug discovery platform Partly sourced | Recursion | L5 | 🌐 | Product revenue not established | 44 | 55 | 25 | 62 | |
AI across Datadog's observability and security platform Partly sourced | Datadog | L3 | 🌐 | Product revenue not established | 43.9 | 75 | 25 | 45 | |
AI customer support agents for enterprise service teams Sourced | Decagon | L5 | 🌐 | Product revenue not established | 43.9 | 75 | 25 | 45 | |
Enterprise AI customer experience agent platform Sourced | Sendbird (센드버드) | L5 | 🇰🇷 | Product revenue not established | 43.9 | 75 | 25 | 45 | |
Enterprise AI contact center Sourced | LG CNS (LG씨엔에스) | L5 | 🇰🇷 | Product revenue not established | 43.9 | 75 | 25 | 45 | |
AI triage and autofix over application errors Partly sourced | Sentry | L3 | 🌐 | Product revenue not established | 43.9 | 75 | 25 | 45 | |
Predictive analytics and anomaly detection over machine data Partly sourced | Splunk | L3 | 🌐 | Product revenue not established | 43.9 | 75 | 25 | 45 | |
AI agents, copilot and quality management in Zendesk Partly sourced | Zendesk | L5 | 🌐 | Product revenue not established | 43.9 | 75 | 25 | 45 | |
DEEPX high-performance edge AI accelerator Partly sourced | DEEPX | L1 | 🇰🇷 | Product revenue not established | 43.8 | 67 | 25 | 50 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
Korean MyData personal-finance AI. Partly sourced | Banksalad (뱅크샐러드) | L5 | 🇰🇷 | Not disclosed | 43.7 | 74 | 25 | 45 | |
AI insurance coverage and claims automation. Partly sourced | Bomapp (보맵) | L5 | 🇰🇷 | Not disclosed | 43.7 | 74 | 25 | 45 | |
Huawei's compute architecture and toolchain for Ascend silicon Partly sourced | Huawei (华为) | L3 | 🇨🇳 | Product revenue not established | 43.7 | 74 | 25 | 45 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
Search, summarisation and review across stored content Partly sourced | Dropbox | L5 | 🌐 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
1.58-bit quantization for AI models on Arm SoCs. Partly sourced | ENERZAi (에너자이) | L3 | 🇰🇷 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
Serverless GPU inference platform for generative media — image, video, audio, and 3D at scale. Partly sourced | fal.ai | L3 | 🌐 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
Compiler and runtime for FuriosaAI NPUs Partly sourced | FuriosaAI (퓨리오사AI) | L3 | 🇰🇷 | Product revenue not established | 43.7 | 74 | 25 | 45 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
Cloud Hancom Office with AI drafting tier Sourced | Hancom (한컴) | L5 | 🇰🇷 | Not disclosed | 43.7 | 74 | 25 | 45 | |
Kakao's AI platform and business messaging stack Partly sourced | Kakao (카카오) | L2 | 🇰🇷 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
Unified proxy and router across LLM providers Partly sourced | LiteLLM | L3 | 🌐 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
Serverless GPU compute for AI workloads Partly sourced | Modal | L3 | 🌐 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
GPU cluster software and virtualization for AI training Partly sourced | Moreh (모레) | L3 | 🇰🇷 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
Writing, Q&A and automation across a Notion workspace Partly sourced | Notion | L5 | 🌐 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
Containerized inference microservices for deploying models on NVIDIA GPUs Sourced | NVIDIA | L3 | 🌐 | Product revenue not established | 43.7 | 74 | 25 | 45 | Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p AWS Trainium |
Open-source LLM inference optimization library for NVIDIA GPUs Sourced | NVIDIA | L3 | 🌐 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
Open-source model serving for multi-framework inference Sourced | NVIDIA | L3 | 🌐 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
Run and fine-tune open models via API Partly sourced | Replicate | L3 | 🌐 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
On-demand and reserved GPU pods for AI training and inference. Sourced | RunPod | L3 | 🌐 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
Autoscaling serverless GPU endpoints for AI inference. Sourced | RunPod | L3 | 🌐 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
AI cross-border remittance with FX hedging and fraud detection. Partly sourced | SentBe (센트비) | L5 | 🇰🇷 | Not disclosed | 43.7 | 74 | 25 | 45 | |
Developer documentation for SiliconFlow model APIs Partly sourced | SiliconFlow | L3 | 🇨🇳 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
Hosted model API and MaaS platform Partly sourced | SiliconFlow | L3 | 🇨🇳 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
Enterprise private deployment for model-serving workloads Partly sourced | SiliconFlow | L3 | 🇨🇳 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
Summarisation, search and workflow automation in Slack Partly sourced | Slack | L5 | 🌐 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
AI co-pilot for Swit work OS. Partly sourced | Swit Technologies (스윗테크놀로지스) | L5 | 🇰🇷 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
On-device AI engine (no GPU needed) Partly sourced | Persona AI | L3 | 🇰🇷 | Product revenue not established | 43.7 | 74 | 25 | 45 | qualcomm-ai apple-core-ml mediatek |
Korean enterprise work OS. Partly sourced | Swit Technologies (스윗테크놀로지스) | L5 | 🇰🇷 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
Inference, fine-tuning and GPU clusters for open models Partly sourced | Together AI | L3 | 🌐 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
VESSL's GPU Liquidity Layer for multi-cloud GPU capacity. Partly sourced | VESSL AI (베슬에이아이) | L3 | 🇰🇷 | Product revenue not established | 43.7 | 74 | 25 | 45 | |
Consumer financial AI assistant built on Ant's finance model Sourced | Ant Group | L5 | 🇨🇳 | Available across Alipay; assistant-specific usage not disclosed | 43.7 | 74 | 25 | 45 | |
ML-driven drug discovery Partly sourced | Insitro | L5 | 🌐 | Product revenue not established | 43.5 | 55 | 25 | 60 | Recursion Tempus Exscientia AI Design |
MFDS-approved AI for osteoporosis screening from routine chest X-rays — addresses 75% undiagnosis rate pre-fracture. Partly sourced | Promedius | L5 | 🇰🇷 | Product revenue not established | 43.5 | 73 | 25 | 45 | ClearRead CT BoneMetrics |
Baidu AI coding assistant Partly sourced | Baidu (百度) | L2 | 🇨🇳 | Product revenue not established | 43.4 | 82 | 25 | 40 | |
Continue's open-source AI coding assistant for IDEs Partly sourced | Continue | L3 | 🌐 | Not disclosed | 43.4 | 82 | 25 | 40 | |
AI assistant and coding agent inside JetBrains IDEs Partly sourced | JetBrains | L3 | 🌐 | Product revenue not established | 43.4 | 82 | 25 | 40 | |
Test generation and code review agents Partly sourced | Qodo | L3 | 🌐 | Product revenue not established | 43.4 | 82 | 25 | 40 | |
Description under review Partly sourced | Tabnine | L3 | 🌐 | Product revenue not established | 43.4 | 82 | 25 | 40 | |
Tencent Cloud AI coding assistant and IDE Partly sourced | Tencent (腾讯) | L2 | 🇨🇳 | Product revenue not established | 43.4 | 82 | 25 | 40 | |
ByteDance's AI development family spanning TRAE IDE and TRAE Work Sourced | ByteDance (字节跳动) | L2 | 🇨🇳 | Product revenue not established | 43.4 | 82 | 25 | 40 | |
Z.ai coding and agent product surface Partly sourced | Zhipu AI (智谱AI) | L2 | 🇨🇳 | Product revenue not established | 43.4 | 82 | 25 | 40 | |
AI character and interactive storytelling service Sourced | Wrtn (뤼튼) | L5 | 🇰🇷 | 550K MAU · est. mobile app, May 2026 (MobileIndex) | 43.3 | 56 | 57.8 | 25 | |
Japanese-focused language model ModelSourced | Sakana AI | L2 | 🌐 | Product revenue not established | 43.3 | 72 | 25 | 45 | - |
Vector search and RAG pipelines for enterprise data PlatformPartly sourced | Dnotitia (디노티시아) | L3 | 🇰🇷 | Product revenue not established | 43.3 | 72 | 25 | 45 | Pinecone PostgreSQL + pgvector |
Korea's largest crowdsourced AI training-data platform Partly sourced | Crowdworks (크라우드웍스) | L3 | 🇰🇷 | Product revenue not established | 43.2 | 77 | 25 | 42 | |
Automotive AI accelerator for ADAS and cockpit AI. Sourced | Bos Semiconductors (보스반도체) | L5 | 🇰🇷 | Product revenue not established | 43.2 | 67 | 25 | 48 | |
Open weights for coding and agentic reasoning ModelSourced | Reflection AI | L2 | 🌐 | Not disclosed | 43.1 | 76 | 30 | 35 | - |
Open-weight reasoning models from OpenAI. ModelSourced | OpenAI | L2 | 🌐 | Not disclosed | 43 | 61 | 45 | 29 | |
Korean humanoid robot trained on real-world human movement data. Sourced | WIRobotics (위로보틱스) | L5 | 🇰🇷 | Product revenue not established | 42.7 | 65 | 25 | 48 | |
Heterogeneous robot fleet-control platform Sourced | Clobot (클로봇) | L5 | 🇰🇷 | Product revenue not established | 42.7 | 74 | 25 | 42 | |
AI image creation with a direct editing workflow AppPartly sourced | Smoretalk (스모어톡) | L5 | 🇰🇷 | Not disclosed | 42.7 | 66 | 42 | 28 | ChatGPT Adobe Firefly |
AI research over patent, chemical and biomedical literature Partly sourced | PatSnap | L5 | 🌐 | Product revenue not established | 42.7 | 71 | 25 | 44 | |
Naver's AI-based product recommendation technology for shopping. Partly sourced | Naver (네이버) | L5 | 🇰🇷 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
AI search, merchandising and recommendations Partly sourced | Algolia | L5 | 🌐 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
AI shopping assistant Sourced | Amazon | L5 | 🌐 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
End-to-end autonomy software for drones and vehicles Partly sourced | Bone AI | L5 | 🇰🇷 | Product revenue not established | 42.5 | 68 | 25 | 45 | anduril helsing skydio |
AI-backed matching inside Hyperconnect's Azar video-chat service. Partly sourced | Hyperconnect | L5 | 🇰🇷 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
Commerce search, personalization and marketing automation Partly sourced | Bloomreach | L5 | 🌐 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
AI video-review collection and merchandising. Partly sourced | Indent (인덴트) | L5 | 🇰🇷 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
Commerce search and discovery optimised for revenue Partly sourced | Constructor | L5 | 🌐 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
Personalization and recommendations for commerce experiences Partly sourced | Dynamic Yield | L5 | 🌐 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
AI across eBay's buying and selling flows Partly sourced | eBay | L5 | 🌐 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
Commerce agents for pricing and product intelligence Partly sourced | Enhans | L5 | 🇰🇷 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
JD Cloud enterprise-agent architecture Partly sourced | JD.com | L5 | 🇨🇳 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
Software and connected-services platform for Kia PBV vehicles. Partly sourced | Kia (기아) | L5 | 🇰🇷 | Product revenue not established | 42.5 | 68 | 25 | 45 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
AI across Klarna's shopping and merchant surfaces Partly sourced | Klarna | L5 | 🌐 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
Game AI for character behaviour and companion NPCs Partly sourced | Krafton (크래프톤) | L5 | 🇰🇷 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
Meituan foundation-model family Partly sourced | Meituan | L5 | 🇨🇳 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
Generative-AI chest X-ray report drafting Sourced | Deepnoid (딥노이드) | L5 | 🇰🇷 | Product revenue not established | 42.5 | 68 | 25 | 45 | |
Conversational AI shopping agent in the Naver Plus Store app. Partly sourced | Naver (네이버) | L5 | 🇰🇷 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
Naver shopping membership and commerce surface Partly sourced | Naver (네이버) | L5 | 🇰🇷 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
Naver's AI-personalized shopping app and commerce surface. Partly sourced | Naver (네이버) | L5 | 🇰🇷 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
Commerce personalization across search, recommendations and segments Partly sourced | Nosto | L5 | 🌐 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
ML-driven Korean lifestyle commerce. Partly sourced | Bucketplace (버킷플레이스) | L5 | 🇰🇷 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
Visual search for fashion e-commerce. Partly sourced | OMNIOUS.AI (옴니어스) | L5 | 🇰🇷 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
AI assistant for merchants Partly sourced | Shopify | L5 | 🌐 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
Unified mobility OS for autonomous Hyundai/Kia vehicles. Partly sourced | 42dot (포티투닷) | L5 | 🇰🇷 | Product revenue not established | 42.5 | 68 | 25 | 45 | |
WATCHA PEDIA's rating-based recommendation engine. Partly sourced | Watcha (왓챠) | L5 | 🇰🇷 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
Chinese social-commerce AI Partly sourced | Xiaohongshu | L5 | 🇨🇳 | Product revenue not established | 42.5 | 73 | 25 | 42 | |
AI cross-channel ad reporting and optimization. Partly sourced | Adriel (애드리엘) | L5 | 🇰🇷 | Product revenue not established | 42.4 | 76 | 25 | 40 | |
AI across SMS and email marketing for retail Partly sourced | Attentive | L5 | 🌐 | Product revenue not established | 42.4 | 76 | 25 | 40 | |
Korean-tuned enterprise RAG platform. Partly sourced | Skelter Labs (스켈터랩스) | L5 | 🇰🇷 | Product revenue not established | 42.4 | 71 | 25 | 43 | |
Predictive targeting and send-time AI in customer engagement Partly sourced | Braze | L5 | 🌐 | Product revenue not established | 42.4 | 76 | 25 | 40 | |
ML-powered native-ad and content-recommendation network. Partly sourced | Dable (데이블) | L5 | 🇰🇷 | Product revenue not established | 42.4 | 76 | 25 | 40 | |
AI drafting and document assistance for Hancom Office environments Sourced | Hancom (한컴) | L5 | 🇰🇷 | Product revenue not established | 42.4 | 71 | 25 | 43 | |
AI across Klaviyo's marketing automation platform Partly sourced | Klaviyo | L5 | 🌐 | Product revenue not established | 42.4 | 76 | 25 | 40 | |
Enterprise AI workspace for search, assistants and agents Partly sourced | Cohere | L5 | 🌐 | Product revenue not established | 42.4 | 71 | 25 | 43 | |
Multi-model agent orchestration behind one API PlatformSourced | Sakana AI | L4 | 🌐 | Product revenue not established | 42.4 | 71 | 25 | 43 | - |
Korean NLP enterprise search engine Sourced | Wisenut (와이즈넛) | L5 | 🇰🇷 | Product revenue not established | 42.4 | 71 | 25 | 43 | |
Document workspace with source-grounded answers and extraction Sourced | Upstage (업스테이지) | L5 | 🇰🇷 | Product revenue not established | 42.4 | 71 | 25 | 43 | |
FuriosaAI vision accelerator for computer-vision workloads Partly sourced | FuriosaAI (퓨리오사AI) | L1 | 🇰🇷 | Product revenue not established | 42.2 | 67 | 25 | 45 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
Xiaomi multimodal models for coding and agents ModelSourced | Xiaomi (小米) | L2 | 🇨🇳 | Not disclosed | 42.1 | 73 | 32 | 32 | - |
SK Telecom's open-weight Korean foundation-model series Sourced | SK Telecom (SK텔레콤) | L2 | 🇰🇷 | Not disclosed | 42 | 66 | 35 | 32 | |
Ambient clinical documentation and coding Partly sourced | Ambience Healthcare | L5 | 🌐 | Product revenue not established | 42 | 74 | 25 | 40 | |
Humanoid robot platform for industrial work Partly sourced | Apptronik | L5 | 🌐 | Product revenue not established | 42 | 55 | 25 | 54 | Datacenter GPU family Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p |
AI logistics management platform Partly sourced | Colosseum | L3 | 🇰🇷 | Product revenue not established | 42 | 74 | 25 | 40 | flexport convoy fourkites |
AI cooking robot that replicates chef-grade grilling with real-time sensing. Sourced | Beyond Honeycomb (비욘드허니컴) | L5 | 🇰🇷 | Product revenue not established | 41.8 | 65 | 25 | 45 | |
Open-source ROS reference mobile robot. Partly sourced | ROBOTIS (로보티즈) | L5 | 🇰🇷 | Global research and education customers | 41.8 | 65 | 25 | 45 | |
Description under review Partly sourced | Boosted.ai | L5 | 🌐 | Product revenue not established | 41.7 | 69 | 25 | 42 | |
Description under review Partly sourced | Dataminr | L5 | 🌐 | Product revenue not established | 41.7 | 69 | 25 | 42 | |
Korean financial data terminal with document and company intelligence Partly sourced | DeepSearch (딥서치) | L5 | 🇰🇷 | Product revenue not established | 41.7 | 69 | 25 | 42 | |
Korea's largest AI robo-advisor by AUM. Partly sourced | Fount (파운트) | L5 | 🇰🇷 | Product revenue not established | 41.7 | 69 | 25 | 42 | |
S&P Global's AI for financial document extraction and linking Partly sourced | S&P Global | L5 | 🌐 | Product revenue not established | 41.7 | 69 | 25 | 42 | |
AI-managed ETFs and asset-management infrastructure Partly sourced | Qraft AI (크래프트테크놀로지스) | L5 | 🇰🇷 | Product revenue not established | 41.7 | 69 | 25 | 42 | |
Description under review Partly sourced | Rogo | L5 | 🌐 | Product revenue not established | 41.7 | 69 | 25 | 42 | |
Z.ai's CogVideoX video-generation model family Partly sourced | Zhipu AI (智谱AI) | L2 | 🇨🇳 | Not disclosed | 41.6 | 72 | 25 | 40 | |
MiniMax video creation application AppSourced | MiniMax | L5 | 🇨🇳 | Not disclosed | 41.6 | 72 | 25 | 40 | |
AI creative suite for video, images and advertising production Sourced | Higgsfield | L5 | 🌐 | 30M+ users (company, Aug 2026; activity definition unspecified) | 41.6 | 72 | 25 | 40 | |
Video, 3D and neural capture generation Partly sourced | Luma AI | L5 | 🌐 | Not disclosed | 41.6 | 72 | 25 | 40 | |
Multimodal video model with native audio ModelSourced | MiniMax | L2 | 🇨🇳 | Not disclosed | 41.6 | 72 | 25 | 40 | |
Pika's AI video generation product Partly sourced | Pika | L5 | 🌐 | Not disclosed | 41.6 | 72 | 25 | 40 | |
AI video-generation platform and API from AIsphere Sourced | PixVerse | L5 | 🇨🇳 | Active-user count not verified; see recorded sources | 41.6 | 72 | 25 | 40 | |
Shengshu's AI video generation model and product Partly sourced | Shengshu Technology | L5 | 🇨🇳 | Not disclosed | 41.6 | 72 | 25 | 40 | |
Alibaba's Wan image and video generation family Partly sourced | Alibaba (阿里巴巴) | L2 | 🇨🇳 | Not disclosed | 41.6 | 72 | 25 | 40 | |
Kakao's open-weight Korean foundation model, moving to agents in KakaoTalk Sourced | Kakao (카카오) | L2 | 🇰🇷 | Not disclosed | 41.5 | 66 | 35 | 31 | |
Conversational AI, voice and AI humans for Korean enterprises Partly sourced | Maum AI (마음AI) | L5 | 🇰🇷 | Product revenue not established | 41.2 | 70 | 25 | 40 | |
LG AI Research's open pathology foundation-model family Sourced | LG AI Research (LG AI연구원) | L2 | 🇰🇷 | Not disclosed | 41.1 | 59 | 42 | 28 | |
AI hiring-match platform for Korean tech talent Sourced | Wanted Lab (원티드랩) | L5 | 🇰🇷 | Product revenue not established | 41 | 69 | 25 | 40 | |
Research toolkit for audio generation and compression Partly sourced | Meta | L3 | 🌐 | Not disclosed | 40.9 | 70 | 25 | 39 | |
Audio AI for spatial sound, loudness and source separation Partly sourced | Gaudio Lab (가우디오랩) | L5 | 🇰🇷 | Product revenue not established | 40.9 | 70 | 25 | 39 | |
Creator AI music composition tool. Partly sourced | Pozalabs (포자랩스) | L5 | 🇰🇷 | Not disclosed | 40.9 | 70 | 25 | 39 | |
MiniMax music-generation product line Partly sourced | MiniMax | L2 | 🇨🇳 | Not disclosed | 40.9 | 70 | 25 | 39 | |
MiniMax speech model and audio product line Partly sourced | MiniMax | L2 | 🇨🇳 | Product revenue not established | 40.9 | 70 | 25 | 39 | ElevenLabs OpenAI TTS Google Cloud Text-to-Speech |
Voice synthesis, conversion, dubbing and speech cleanup Partly sourced | Supertone (수퍼톤) | L5 | 🇰🇷 | Product revenue not established | 40.9 | 70 | 25 | 39 | |
AI music generation with editing controls Partly sourced | Udio | L5 | 🌐 | Not disclosed | 40.9 | 70 | 25 | 39 | |
Korean AI agent for nuanced search and deep research Partly sourced | ESTsoft (이스트소프트) | L5 | 🇰🇷 | Not disclosed | 40.8 | 68 | 25 | 40 | Perplexity ChatGPT Liner |
AI research and briefing assistant for professionals Partly sourced | Saltlux (솔트룩스) | L5 | 🇰🇷 | Enterprise | 40.8 | 68 | 25 | 40 | Perplexity Glean coveo |
AI-generated search briefings embedded in Naver Search Sourced | Naver (네이버) | L2 | 🇰🇷 | Not disclosed | 40.8 | 68 | 25 | 40 | |
Naver's browser-based AI voice and video dubbing service Sourced | Naver (네이버) | L2 | 🇰🇷 | Product revenue not established | 40.7 | 69 | 25 | 39 | |
Korean STT, transcription, and meeting AI. Partly sourced | ActionPower (액션파워) | L5 | 🇰🇷 | Product revenue not established | 40.7 | 69 | 25 | 39 | |
Machine translation, writing assistance and speech Partly sourced | DeepL | L5 | 🌐 | Product revenue not established | 40.7 | 69 | 25 | 39 | |
Hybrid human + AI translation platform. Partly sourced | Flitto (플리토) | L5 | 🇰🇷 | Product revenue not established | 40.7 | 69 | 25 | 39 | |
AI translation and discovery for Korean webtoons and web novels Partly sourced | RIDI | L5 | 🇰🇷 | Product revenue not established | 40.7 | 69 | 25 | 39 | |
AI localization pipeline for subtitles and dubbing Partly sourced | Voithru (보이스루) | L5 | 🇰🇷 | Product revenue not established | 40.7 | 69 | 25 | 39 | |
Bimanual robot manipulation foundation model Sourced | Config Intelligence (컨피그인텔리전스) | L2 | 🇰🇷 | Product revenue not established | 40.5 | 59 | 25 | 45 | |
AI drug discovery with 230M+ data entries. Partly sourced | Pharos iBio (파로스아이바이오) | L5 | 🇰🇷 | Product revenue not established | 40.4 | 55 | 25 | 48 | |
Life-sciences research model for biology and drug-discovery workflows ModelSourced | OpenAI | L2 | 🌐 | Product revenue not established | 40.4 | 55 | 25 | 48 | |
Self-serve SaaS AI drug-discovery workbench Sourced | HITS (히츠) | L5 | 🇰🇷 | Product revenue not established | 40.4 | 55 | 25 | 48 | |
AI CGM-based blood-glucose management app Sourced | Kakao Healthcare (카카오헬스케어) | L5 | 🇰🇷 | Product revenue not established | 40.4 | 66 | 25 | 40 | |
ML and LLM observability with open-source tracing Partly sourced | Arize AI | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
Evaluation and observability for LLM product teams Partly sourced | Braintrust | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
Conversation analytics for LLM product teams Partly sourced | Coxwave (콕스웨이브) | L5 | 🇰🇷 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
AI governance, policy and regulatory readiness Partly sourced | Credo AI | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
Training-data production and LLM evaluation for Korean AI builders Partly sourced | Datumo (셀렉트스타) | L3 | 🇰🇷 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
AI model evaluation and monitoring Partly sourced | Datumo (셀렉트스타) | L3 | 🇰🇷 | Product revenue not established | 40.1 | 68 | 25 | 38 | Scale Data Engine Arize arthur-ai |
Open-source LLM evaluation and red-teaming framework Partly sourced | DeepEval | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
Evidently's open-source ML and LLM evaluation and monitoring Partly sourced | Evidently AI | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
AI-assisted 3D sketching. Partly sourced | Sketchsoft (스케치소프트) | L5 | 🇰🇷 | Not disclosed | 40.1 | 66 | 25 | 39 | |
Model monitoring, explainability and LLM guardrails Partly sourced | Fiddler AI | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
Evaluation, monitoring and guardrails for LLM applications Partly sourced | Galileo | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
Open-source testing and vulnerability scanning for AI models Partly sourced | Giskard | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
Guardrails AI's validation framework for LLM output Partly sourced | Guardrails AI | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
Great Expectations data-quality validation, open-source and cloud Partly sourced | Great Expectations | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
Runtime guardrails and red-teaming for LLM applications Partly sourced | Lakera | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
Open-source LLM engineering and observability platform Partly sourced | Langfuse | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
LLM observability, tracing, and evaluation platform Partly sourced | LangChain | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
Generate, edit and export persistent 3D worlds AppSourced | World Labs | L5 | 🌐 | Not disclosed | 40.1 | 66 | 25 | 39 | - |
3D asset generation from text and images Partly sourced | Meshy | L5 | 🌐 | Not disclosed | 40.1 | 66 | 25 | 39 | |
Generative AI for 3D character motion from text, voice, or video. Partly sourced | NationA (네이션에이) | L5 | 🇰🇷 | Product revenue not established | 40.1 | 66 | 25 | 39 | |
AI auto-tracking smartphone camera mount. Partly sourced | 3i Inc. (쓰리아이) | L5 | 🇰🇷 | Not disclosed | 40.1 | 66 | 25 | 39 | |
AI motion capture in the browser. Partly sourced | Plask (플라스크) | L5 | 🇰🇷 | Not disclosed | 40.1 | 66 | 25 | 39 | |
Open-source prompt testing and LLM red-teaming Partly sourced | Promptfoo | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
Open-source evaluation framework for RAG systems Partly sourced | Ragas | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
Experiment tracking and LLM observability for ML teams Partly sourced | Weights & Biases | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
Data and ML monitoring with privacy-preserving telemetry Partly sourced | WhyLabs | L3 | 🌐 | Product revenue not established | 40.1 | 68 | 25 | 38 | |
AI21 hybrid SSM-Transformer model family Partly sourced | AI21 Labs | L2 | 🌐 | Product revenue not established | 39.8 | 74 | 25 | 34 | |
ModelBest's efficient multimodal model family for edge deployment ModelSourced | ModelBest | L2 | 🇨🇳 | Not disclosed | 39.8 | 74 | 25 | 34 | |
Open-weight Korean LLM trained from scratch. Sourced | Motif Technologies (모티프테크놀로지스) | L2 | 🇰🇷 | Not disclosed | 39.8 | 74 | 25 | 34 | |
StepFun's Step foundation-model family ModelSourced | StepFun | L2 | 🇨🇳 | Not disclosed | 39.8 | 74 | 25 | 34 | |
Korean diffusion-transformer LLM. Sourced | Trillion Labs (트릴리언랩스) | L2 | 🇰🇷 | Not disclosed | 39.8 | 74 | 25 | 34 | |
AI-powered UI generation by Vercel Partly sourced | Vercel | L3 | 🌐 | Not disclosed | 39.8 | 72 | 25 | 35 | |
01.AI open and commercial foundation-model family Sourced | 01.AI (零一万物) | L2 | 🇨🇳 | Not disclosed | 39.8 | 74 | 25 | 34 | |
Voice AI assistant Sourced | Amazon | L5 | 🌐 | 500M+ Alexa-enabled devices sold (installed base, not AI-product users) | 39.7 | 78 | 25 | 32 | |
Upstage's messenger-based AI assistant with document OCR Sourced | Upstage (업스테이지) | L2 | 🇰🇷 | Not disclosed | 39.7 | 78 | 25 | 32 | |
ByteDance AI assistant and model-facing consumer app AppSourced | ByteDance (字节跳动) | L5 | 🇨🇳 | Active-user count not verified; see recorded sources | 39.7 | 78 | 25 | 32 | ERNIE (文心) ChatGPT |
All-in-one AI workspace and agentic research platform. Sourced | Genspark | L5 | 🌐 | Not disclosed | 39.7 | 78 | 25 | 32 | |
SK Telecom's Korean voice-assistant and device ecosystem Sourced | SK Telecom (SK텔레콤) | L2 | 🇰🇷 | Not disclosed | 39.7 | 78 | 25 | 32 | |
iFLYTEK's Spark assistant and model family Partly sourced | iFlytek (科大讯飞) | L5 | 🇨🇳 | Not disclosed | 39.7 | 78 | 25 | 32 | |
Mistral AI assistant and long-horizon work agent, formerly Le Chat Partly sourced | Mistral AI | L2 | 🌐 | Not disclosed | 39.7 | 78 | 25 | 32 | |
Tencent AI assistant Partly sourced | Tencent (腾讯) | L2 | 🇨🇳 | Not disclosed | 39.7 | 78 | 25 | 32 | |
Framework for programming and optimising LLM pipelines Partly sourced | DSPy | L3 | 🌐 | Not disclosed | 39.6 | 71 | 25 | 35 | |
Connect agents to documents and enterprise data Partly sourced | LlamaIndex | L3 | 🌐 | Not disclosed | 39.6 | 71 | 25 | 35 | |
Framework for customizing and evaluating models on NVIDIA infrastructure Sourced | NVIDIA | L3 | 🌐 | Enterprise model customization framework; user counts not disclosed | 39.6 | 71 | 25 | 35 | Instinct GPU MI300X/MI350/MI400 Cloud TPU v5e/v5p AWS Trainium |
China's largest open-source deep-learning framework and model platform Sourced | Baidu (百度) | L3 | 🇨🇳 | Active-user count not verified; see recorded sources | 39.6 | 71 | 25 | 35 | |
TypeScript toolkit and gateway for building AI applications Partly sourced | Vercel | L3 | 🌐 | Not disclosed | 39.6 | 71 | 25 | 35 | |
AI loan-prediction and marketplace. Sourced | Finda (핀다) | L5 | 🇰🇷 | Product revenue not established | 39.5 | 65 | 25 | 38 | |
Z.ai vision and image-generation model surface Partly sourced | Zhipu AI (智谱AI) | L2 | 🇨🇳 | Not disclosed | 39.4 | 70 | 25 | 35 | |
AI image generation inside a stock-asset library Partly sourced | Freepik | L5 | 🌐 | Not disclosed | 39.4 | 70 | 25 | 35 | |
Persistent agent runtime with reusable skills Agent harnessSourced | Nous Research | L4 | 🌐 | Not disclosed | 39.4 | 73 | 30 | 28 | |
Ideogram's AI image generator with reliable in-image text Partly sourced | Ideogram | L5 | 🌐 | Not disclosed | 39.4 | 70 | 25 | 35 | |
Real-time generative image and video canvas Partly sourced | Krea | L5 | 🌐 | Not disclosed | 39.4 | 70 | 25 | 35 | |
Image generation with fine-tuned models for game and product art Partly sourced | Leonardo AI | L5 | 🌐 | Not disclosed | 39.4 | 70 | 25 | 35 | |
Meitu's imaging and video AI across photo, video and design Partly sourced | Meitu (美图) | L5 | 🇨🇳 | Not disclosed | 39.4 | 70 | 25 | 35 | |
Photorealistic AI images Partly sourced | L5 | 🌐 | Enterprise | 39.4 | 70 | 25 | 35 | dall-e Midjourney Stable Diffusion | |
Recraft's AI design and vector-graphics generator Partly sourced | Recraft | L5 | 🌐 | Not disclosed | 39.4 | 70 | 25 | 35 | |
Established image-model family for self-hosted and API generation Sourced | Stability AI | L5 | 🌐 | Not disclosed | 39.4 | 70 | 25 | 35 | |
AI document scanner for mobile. Partly sourced | VoyagerX (보이저엑스) | L5 | 🇰🇷 | Product revenue not established | 39.4 | 66 | 25 | 37 | |
Meitu AI image creation product Partly sourced | Meitu (美图) | L5 | 🇨🇳 | Not disclosed | 39.4 | 70 | 25 | 35 | |
Extensible agent harness and coding CLI Agent harnessSourced | Earendil | L4 | 🌐 | Not disclosed | 39.3 | 73 | 32 | 26 | - |
Plan, generate and edit visual content in one studio AppPartly sourced | RECON Labs (리콘랩스) | L5 | 🇰🇷 | Not disclosed | 38.8 | 73 | 25 | 32 | Runway Adobe Firefly |
Open-weight MoE reasoning model for evaluation and adaptation ModelPartly sourced | Trillion Labs (트릴리언랩스) | L2 | 🇰🇷 | Product revenue not established | 38.8 | 73 | 25 | 32 | Qwen (通义千问) Gemma |
BHSN's all-in-one legal AI platform for enterprise legal and business teams. Partly sourced | BHSN (비에이치에스엔) | L5 | 🇰🇷 | Product revenue not established | 38.7 | 66 | 25 | 35 | |
Korean case-law database with AI search and analysis Partly sourced | LBox | L5 | 🇰🇷 | Product revenue not established | 38.7 | 66 | 25 | 35 | |
AI lesson planning and teaching tools for K-12 educators Partly sourced | MagicSchool AI | L5 | 🌐 | Product revenue not established | 38.5 | 65 | 25 | 35 | |
Adaptive tutoring system for Chinese K-12 after-school learning Partly sourced | Squirrel AI | L5 | 🇨🇳 | Product revenue not established | 38.5 | 65 | 25 | 35 | |
AI avatar video creation platform Partly sourced | DeepBrain AI (딥브레인AI) | L5 | 🇰🇷 | Product revenue not established | 38.3 | 66 | 25 | 34 | Synthesia HeyGen D-ID Agents |
Generative virtual-human platform behind Eternity Sourced | Pulse9 (펄스나인) | L5 | 🇰🇷 | Product revenue not established | 38.3 | 66 | 25 | 34 | |
AI virtual-idol platform for K-pop. Partly sourced | Galaxy Corporation (갤럭시코퍼레이션) | L5 | 🇰🇷 | Product revenue not established | 38.3 | 66 | 25 | 34 | |
Real-time AI avatar and digital-human studio. Partly sourced | Klleon (클레온) | L5 | 🇰🇷 | Product revenue not established | 38.3 | 66 | 25 | 34 | |
Virtual-human video generation. Sourced | ESTsoft (이스트소프트) | L5 | 🇰🇷 | Product revenue not established | 38.3 | 66 | 25 | 34 | |
Tencent digital-human and avatar product surface Partly sourced | Tencent (腾讯) | L2 | 🇨🇳 | Product revenue not established | 38.3 | 66 | 25 | 34 | |
AI across Alibaba's DingTalk workplace suite Partly sourced | DingTalk | L5 | 🇨🇳 | Product revenue not established | 38.1 | 67 | 25 | 33 | |
AI across ByteDance's Feishu/Lark workplace suite Partly sourced | Feishu | L5 | 🇨🇳 | Product revenue not established | 38.1 | 67 | 25 | 33 | |
A compact device for local language-model use Sourced | Dnotitia (디노티시아) | L3 | 🇰🇷 | Product revenue not established | 38.1 | 63 | 25 | 35 | Mac Studio / mini Standard PC + local runtime |
AI video note-taking with transcript Q&A. Partly sourced | Bebridge (비브릿지) | L5 | 🇰🇷 | Product revenue not established | 38.1 | 67 | 25 | 33 | |
AI call-recording and transcription app Sourced | Return Zero (리턴제로) | L5 | 🇰🇷 | Product revenue not established | 38.1 | 67 | 25 | 33 | Naver Clova Note Otter Daglo (다글로) |
Health-behavior AI agent recognized at CES 2026. Partly sourced | WELT (웰트) | L5 | 🇰🇷 | Not disclosed | 37.7 | 65 | 25 | 33 | |
Adaptive learning across subjects with teacher analytics Partly sourced | Century Tech | L5 | 🌐 | Product revenue not established | 37.7 | 65 | 25 | 33 | |
AI coaching and course authoring inside Coursera Sourced | Coursera | L5 | 🌐 | Not disclosed | 37.7 | 65 | 25 | 33 | |
AI-assisted coding education and cloud learning platform Partly sourced | Elice (엘리스) | L5 | 🇰🇷 | Product revenue not established | 37.7 | 65 | 25 | 33 | |
AI study tools over Quizlet's study-set corpus Partly sourced | Quizlet | L5 | 🌐 | Not disclosed | 37.7 | 65 | 25 | 33 | |
Output-focused K-12 writing and speaking learning Partly sourced | Socra AI | L5 | 🇰🇷 | Product revenue not established | 37.7 | 65 | 25 | 33 | |
ESL learning with licensed real-world content Partly sourced | Socra AI | L5 | 🇰🇷 | Not disclosed | 37.7 | 65 | 25 | 33 | |
Beginner-friendly ESL speaking practice Partly sourced | Socra AI | L5 | 🇰🇷 | Not disclosed | 37.7 | 65 | 25 | 33 | |
Ringle's post-lesson AI speech analysis feature. Partly sourced | Ringle | L5 | 🇰🇷 | Not disclosed | 37.7 | 65 | 25 | 33 | |
AI assistant and learning platform over company knowledge Partly sourced | Sana Labs | L5 | 🌐 | Product revenue not established | 37.7 | 65 | 25 | 33 | |
Financial education with market-data-powered AI guidance Partly sourced | Socra AI | L5 | 🇰🇷 | Not disclosed | 37.7 | 65 | 25 | 33 | |
AI disease-risk prediction from health-checkup data Sourced | Selvas AI (셀바스AI) | L5 | 🇰🇷 | Product revenue not established | 37.4 | 60 | 25 | 35 | Lydia AI Klarity Health AITRICS VC |
Open multimodal models built for customization ModelSourced | Thinking Machines Lab | L2 | 🌐 | Not disclosed | 36.7 | 68 | 25 | 29 | - |
Small time-series models for probabilistic forecasting ModelSourced | Mistral AI | L2 | 🌐 | Product revenue not established | 36.5 | 62 | 28 | 28 | - |
AI writing workspace connected to research. Sourced | Liner (라이너) | L5 | 🇰🇷 | Yonsei access agreement · Mar 2026; module usage undisclosed | 35.4 | 53 | 42 | 20 | |
Typed decisions and confidence estimates for software ModelSourced | TypeSafe AI | L2 | 🌐 | Product revenue not established | 35 | 61 | 28 | 25 | - |
Healthcare-tuned open model family from Google Partly sourced | L2 | 🌐 | Not disclosed | 34.6 | 59 | 25 | 28 | ||
Open vision-language model family from Google Partly sourced | L2 | 🌐 | Not disclosed | 34.6 | 59 | 25 | 28 | ||
Therapeutics and drug-discovery open model family from Google Partly sourced | L2 | 🌐 | Not disclosed | 34.6 | 59 | 25 | 28 | ||
NC AI's game/content foundation-AI platform with a physical-AI pivot Sourced | NCSOFT (엔씨소프트) | L2 | 🇰🇷 | Product revenue not established | 34.6 | 59 | 25 | 28 | |
EXAONE scientific discovery model effort Sourced | LG AI Research (LG AI연구원) | L2 | 🇰🇷 | Not disclosed | 33.6 | 48 | 22 | 36 | |
4nm LLM-inference accelerator on LPU architecture Sourced | HyperAccel (하이퍼엑셀) | L1 | 🇰🇷 | Product revenue not established | 32.8 | 49 | 18 | 40 | |
Chat with AI characters Sourced | Character AI | L5 | 🌐 | Active-user count not verified; see recorded sources | 32.7 | 56 | 25 | 25 | |
MiniMax AI companion and character product Sourced | MiniMax | L2 | 🇨🇳 | Not disclosed | 32.7 | 56 | 25 | 25 | |
Brand-conditioned marketing content generation Partly sourced | Jasper | L5 | 🌐 | Product revenue not established | 29.8 | 53 | 25 | 20 | |
AI features across the Shopify merchant surface Partly sourced | Shopify | L5 | 🌐 | Product revenue not established | 29.8 | 53 | 25 | 20 | |
AI rewriting and writing assistance Partly sourced | AI21 Labs | L5 | 🌐 | Not disclosed | 29.8 | 53 | 25 | 20 | |
Research image-and-text base model with downloadable weights ModelSourced | Adept AI | L2 | 🌐 | Not disclosed | 26.6 | 42 | 25 | 18 | - |
Early-stage Korean general-purpose humanoid platform. Partly sourced | Holiday Robotics (홀리데이로보틱스) | L5 | 🇰🇷 | Product revenue not established | 25.2 | 36 | 12 | 37 |