Reasoning Models

The OpenAI reasoning models, the o1, o3 models, DeepSeek R1 or Anthropic Sonnet 3.7, are models that are optimized for reasoning tasks.

script({
    model: "openai:o1",
})

Tip

You can experiment with these models on Github Models as well but the context window is quite small.

script({
    model: "github:openai/o3-mini",
})

Model Alias

The reasoning and reasoning-small model aliases are available for reasoning models.

script({
    model: "openai:reasoning",
})

or

genaiscript run ... -p openai -m reasoning

Reasoning, thinking

GenAIScript automatically extracts the thinking/reasoning content of the LLM responses.

Reasoning effort

The reasoning effort parameter can be set to low, medium, or high.

• configured with the reasoningEffort parameter

script({
    model: "openai:o3-mini"
    reasoningEffort: "high"
})

• as a tag to the model name

script({
    model: "openai:o3-mini:high",
})

For Anthropic Sonnet 3.7, the reasoning efforts are mapped to the following budget_token values:

• low: 2048
• medium: 4096
• high: 16384

Limitations

• o1-preview, o1-mini do not support streaming
• o1 models do not support tool calling so GenAIScript uses fallback tools.

Advice on prompting

OpenAI provides an extensive advice on prompting reasoning models.