Provider pricing

OctoAI AI API pricing

OctoAI (acquired by NVIDIA in 2024) offers a hosted inference platform for open-source LLMs, image, and audio models with per-token and per-second billing. Customers can also bring custom checkpoints and serve them on optimized NVIDIA infrastructure.

Founded2020
HeadquartersSeattle, WA
Tracked modelsDirectory

Best for

Teams that want NVIDIA-optimized inference for open and custom models.

Strengths

  • Backed by NVIDIA's GPU and inference stack
  • Per-token pricing on Llama, Mixtral, and Qwen
  • Custom checkpoint hosting with autoscaling
  • Image and audio model APIs alongside LLMs