FriendliAI
High-performance inference cloud for serving frontier AI models at scale.
About
FriendliAI is an inference platform founded by Byung-Gon Chun, the Seoul National University professor whose Orca paper introduced continuous batching, the technique at the heart of vLLM. Its PeriFlow engine serves large language models with patented batching technology aimed at maximum throughput and low cost.
The platform offers PeriFlow Cloud for managed deployment with monitoring and auto-scaling, and InferenceSense, which lets neocloud operators monetize idle GPU cycles with paid inference workloads. In 2026 the company partnered with Samsung SDS to run frontier open-weight models on Samsung Cloud Platform's NVIDIA B300 infrastructure.
What it can do
- PeriFlow high-throughput serving engine
- PeriFlow Cloud managed deployment
- InferenceSense idle-GPU monetization
- Day-0 support for frontier open-weight models
Setup
Sign up on the FriendliAI website, upload a model checkpoint to PeriFlow Cloud, deploy, and call the OpenAI-compatible inference endpoints.
Prerequisites
An account; a model checkpoint to deploy or use of the hosted frontier models.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 116; Trending score 78/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.