| Provider | Models | Notes |
|---|---|---|
| OpenAI | gpt-5.1, gpt-5.1-mini, gpt-5.1-nano, gpt-5, gpt-5-mini, gpt-5-nano, gpt-4o, gpt-4o-mini | Policy models for GEPA/MIPRO. gpt-5-pro is rejected (cost guardrail). |
| Groq | openai/gpt-oss-20b, openai/gpt-oss-120b, meta/llama-3.3-70b-versatile, qwen/qwen3-32b, moonshotai/kimi-k2-0905, meta/llama-guard-4-12b, meta/llama-3.1-8b-instant | Configure provider credentials in the backend. |
gemini-3-pro-preview, gemini-3-pro-preview-gt200k, gemini-2.5-pro, gemini-2.5-pro-gt200k, gemini-2.5-flash, gemini-2.5-flash-lite | Use as policy models where latency/cost fit. |
Supported Models
Supported Models for Prompt Optimization
Prompt optimization (GEPA and MIPRO) supports the following policy models:
See Prompt Optimization Configuration for policy/mutation/meta model details.***