Provider pricing

Friendli AI AI API pricing

Friendli AI runs Llama 4, Qwen3, DeepSeek, and Mixtral on its custom Friendli Engine, which uses iteration batching to deliver 2–4× higher throughput than vanilla vLLM. Serverless endpoints offer per-token pricing, and dedicated endpoints provide guaranteed capacity with autoscaling.

Founded2021
HeadquartersRedwood City, CA
Tracked modelsDirectory

Best for

High-throughput production workloads on open-weight models.

Strengths

  • Friendli Engine: 2–4× higher throughput vs vLLM via iteration batching
  • Serverless and dedicated endpoints on the same platform
  • Quantization-aware serving with no quality degradation
  • OpenAI-compatible API with streaming and tool use