Platform

Models we support

The model families and runtimes we work with, and how we pick.


We're model-agnostic and pick the right model for each task and budget — you're never locked to one vendor.

Open / self-hostable

  • Llama
  • Mistral
  • Qwen
  • DeepSeek
  • Gemma
  • Phi
  • GLM
  • Kimi

API providers

  • OpenAI GPT
  • Anthropic Claude
  • Google Gemini
  • xAI Grok
  • Cohere

Runtime & infrastructure

  • vLLM
  • Ollama
  • llama.cpp
  • Hugging Face TGI
  • RunPod

For a given use case we'll benchmark a couple of options and recommend the best balance of quality, speed, and cost.