Supported Models

Refrase optimizes prompts for 11 models across 11 providers.

Anthropic

Claude Sonnet 4.6

Anthropic+12%

claude

XML tag structuringRole extractionOutput format guidance

Google

Google

gemini

OpenAI

OpenAI+8%

openai

Grounding rulesJSON reinforcementReasoning hints

Alibaba

Alibaba+15%

qwen

Thinking mode controlEnglish enforcement

DeepSeek

DeepSeek+10%

deepseek

Self-verificationJSON reinforcement

Mistral

Mistral Large 3

Mistral+7%

mistral

Step-by-step for analysisJSON reinforcement

Meta

Meta+9%

llama

Grounding rulesJSON reinforcement

Moonshot

Moonshot+11%

kimi

Source groundingEnglish enforcement

Z.AI

Z.AI+13%

glm

Nested object fixEnglish enforcement

NVIDIA

NVIDIA+6%

nemotron

JSON reinforcement

MiniMax

MiniMax+8%

minimax

Self-verificationJSON reinforcement