FriendliAI
Enterprise LLM inference at up to 90% lower GPU cost — built on the ORCA research that invented continuous batching
Overview
FriendliAI is an inference platform founded in 2021 as a Seoul National University spinout, commercializing the ORCA paper (OSDI 2022) that introduced continuous batching. It serves 580,000+ open-weight models via serverless API, dedicated GPU, or self-hosted Container deployment.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 11 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- No infra to manage
- OpenAI-compatible API
- Per-model token pricing
- A100 to B200
- No cold-start fees
- Predictable capacity
- Full data sovereignty
- On-prem or VPC
- Custom rate limits
- Reserved capacity
- Dedicated support
Use cases
What you can produce with FriendliAI
- Substantial GPU serving cost reductions versus baseline inference deployments, as reported across enterprise deployments
- OpenAI-compatible REST API allowing drop-in substitution of base URL with no other code changes
- Access to over 570,000 Hugging Face open-weight text, vision, audio, and multi-modal models via Serverless Endpoints, including EXAONE 4.0 at launch
- Dedicated GPU endpoints on A100, H100, H200, and B200 GPUs billed per second with enterprise SLA
- Friendli Container deployable on AWS EKS, on-prem, or air-gapped private cloud with SOC 2 Type II and HIPAA compliance
- Usage-tiered rate limits scaling from 100 RPM (Tier 1) to 10,000 RPM (Tier 4) based on cumulative spend
- Structured output, function/tool calling, JSON mode, vision, and audio (Whisper) inference from a unified API
ASEAN Perspective
FriendliAI in Southeast Asia
FriendliAI has meaningful APAC roots and traction: founded as a Seoul National University spinout, it maintains a Seoul operational hub and counts South Korean majors SK Telecom, Scatter Lab (active across Korea and Japan), and Upstage as named customers. The exclusive API launch partnership for LG AI Research's EXAONE 4.0 model is a strong signal of its position as preferred inference partner for Korean AI labs. The $20M seed extension (August 2025) included Korean institutional investors KDB and KB Securities alongside US VCs, and the company has explicitly flagged Asia expansion alongside North America as a use of proceeds. ASEAN-headquartered teams will find a vendor with demonstrated Korean-market depth, OpenAI-compatible APIs, and a cost-reduction pitch that translates well to GPU-cost-sensitive emerging market AI deployments.
FriendliAI's edge is real research pedigree — its founders wrote the ORCA paper that gave the industry continuous batching, and the custom-built inference engine that grew out of it delivers cost and throughput numbers that are hard to wave away: SK Telecom cites 5x throughput and 3x cost savings, Scatter Lab scaled past a billion monthly interactions on Friendli Container. The catalogue now covers 580,000+ Hugging Face models via an OpenAI-compatible API, and it's the exclusive launch partner for LG's EXAONE 4.0. The catch is scale: with roughly 25-30 large clients and under 60 employees as of mid-2025, this is still a seed-stage vendor. Revenue growth figures (6-7x for 2025) are self-reported, there's no instant free sign-up — the credit programme needs approval — and teams with strict multi-region SLAs should verify capacity directly with sales.
What people say
FriendliAI's dedicated GPU pricing starts at $2.90/hour for an A100 and tops out around $8.90/hour for a B200 — competitive against renting the same hardware raw, and the company's core claim is that its inference engine, built from the ORCA continuous-batching research (OSDI 2022, the technique vLLM later adopted), squeezes 50-90% cost savings and multi-fold throughput gains out of open-weight models versus a generic serving stack.
Production evidence backs some of that up: SK Telecom reports 5x throughput and 3x cost savings for enterprise AI agents, Scatter Lab runs its Zeta chat service past a billion monthly interactions on Friendli Container across Korea, Japan, and the US, and FriendliAI is the exclusive API launch partner for LG AI Research's EXAONE 4.0. The catalogue now spans 580,000+ deployable Hugging Face models across text, vision, audio, and multimodal, reachable through an OpenAI-compatible API.
What holds it back from an easy recommendation is size and transparency. This is still a roughly 25-30-large-client, under-60-employee company that closed a $20M seed extension in August 2025 — its 6-7x revenue growth projection for 2025 is a company number, not an audited one. There's no self-serve free tier; the $10K credit programme requires an application. Good fit for teams that have outgrown closed-model APIs and want cost-efficient open-weight serving with an actual research team behind the engine.
Summary of public user & expert reviews, compiled by RECATOOLS.
Notable facts
- The ORCA paper published by FriendliAI's founding team at OSDI 2022 introduced continuous batching — the scheduling technique that now powers vLLM and most major open-source inference engines, meaning the entire industry runs on an idea from this team.
- Scatter Lab used Friendli Container to scale its Zeta conversational AI to over 1 billion monthly interactions across Korea, Japan, and the US — reaching financial breakeven in the process.
- FriendliAI supports over 430,000 models from Hugging Face and acted as the exclusive API launch provider for LG AI Research's 32-billion-parameter EXAONE 4.0 model.
- CEO Byung-Gon Chun holds a PhD from UC Berkeley and an MS from Stanford, yet returned to Seoul National University as a professor before spinning out FriendliAI — a rare case of a top Korean university producing a Silicon Valley-headquartered AI infrastructure company.
Frequently asked questions
About this listing
This entry was compiled from publicly available data including FriendliAI's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with FriendliAI unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to FriendliAI directly →
Spotted something out of date? Suggest an update →
More in Code & Dev Tools
