01
FriendliAI · korea · Strength 51.9
Managed serverless model APIs
Dedicated endpoints where GPU efficiency and predictable availability matter
Customer count and contract pricing are not disclosed
Buyer guide
Inference, cluster operations, and multimodal APIs built by Korean or Korean-founded teams.
Selection principle
Compare deployment control, workload evidence, and switching cost before comparing headline speed.
Updated 2026-09-04
01
FriendliAI · korea · Strength 51.9
Managed serverless model APIs
Dedicated endpoints where GPU efficiency and predictable availability matter
Customer count and contract pricing are not disclosed
02
Lablup · korea · Strength 56
Multi-tenant accelerator orchestration
Teams that need a self-hosted path as well as a managed option
Deployment scale is described qualitatively
03
Twelve Labs · korea · Strength 51
Semantic video search
Video question answering and event understanding through an API
Usage-based cost depends on video volume
04
VESSL AI · korea · Strength 43.7
Multi-cloud GPU provisioning
Spot, on-demand, and reserved capacity
No product-specific adoption evidence is recorded
These are explainable shortlists, not paid placements. Inclusion uses the criteria above and the evidence attached to each catalog row. Missing information is shown as a limitation rather than silently estimated.