Models

Models

All models are accessed through the LiteLLM proxy at api.stilosolutions.com. Use them exactly like the OpenAI API — just swap the base URL and use your LiteLLM key.

Available models

Model nameTypeBest for
qwen3-coderLocal (GPU)Code generation, review, debugging
qwen3-generalLocal (GPU)Research, writing, analysis, general tasks
claude-sonnetCloud (Anthropic)Complex reasoning, nuanced writing
claude-opusCloud (Anthropic)Hardest tasks where quality is critical
claude-haikuCloud (Anthropic)Fast summaries, cheap high-volume tasks

Local models are free (electricity only). Cloud models are billed to the company Anthropic account — use them when local quality isn't sufficient.

Connecting your tools

# In your shell / .bashrc / .env
export OPENAI_API_KEY=$(cat ~/.config/litellm/key)
export OPENAI_BASE_URL=https://api.stilosolutions.com/v1

Any OpenAI-compatible library (openai, litellm, crewai, langchain) works with these settings.

Fallback routing

If a local model is unreachable, the proxy automatically falls back:

  • qwen3-coderclaude-sonnet
  • qwen3-generalclaude-haiku

You don't need to handle this in your code.