Maestro
The model board
The board is the set of models Maestro uses most for each kind of work. It changes when a better model ships or a price moves.
Twelve models are on the board today. Behind them are 300+ AI model endpoints across 14 leading model providers: OpenAI, Anthropic, Google, xAI, DeepSeek, Qwen, GLM, Kimi, MiniMax, Mistral, NVIDIA, Cohere, Perplexity and more.
| Kind of work | Model today |
|---|---|
| Specs | Fable |
| Drafting, planning | GPT |
| Hard planning | Astra |
| Code, summaries | DeepSeek |
| Testing | Sonnet |
| Review | Opus |
| Research | Grok |
| Images | Gemini |
| Translation | GLM |
| Extraction | Qwen |
| Cleanup | Llama |
| Tagging | Mistral |
When a better model ships
A new model joins the board once it proves itself on real tasks. Your AI team uses it from the next task, with nothing to change on your side.
The board is a starting point
The board is where most tasks go today, not a fixed list. For any one task, Maestro can route to any model behind the board when it does that task better or for less. And you can pin a model to a kind of work.
Was this page helpful?Thank you. Tell us more at apply@zbstack.xyz.