modelparams.dev

LLM parameter glossary

28 parameters appear across the catalog. This page defines each one, grouped by what it controls, and notes its type and how many models expose it. Definitions come from the same community-maintained data as the JSON API.

Length

Max tokens max_tokens integer · 52 models

Maximum number of output tokens the model may generate.

Supported by: Anthropic, DeepSeek, Mistral, OpenAI

Max tokens max_completion_tokens integer · 13 models

Maximum number of output tokens the model may generate.

Supported by: OpenAI

Stop sequence stop string · 13 models

Stops generation when this string is detected.

Supported by: Mistral

Max output tokens generationConfig.maxOutputTokens integer · 4 models

Maximum number of tokens to include in a response candidate.

Supported by: Google

Sampling

Temperature temperature number · 49 models

Controls randomness. Lower values make outputs more focused; higher values make them more varied.

Supported by: Anthropic, DeepSeek, Mistral, OpenAI

Top P top_p number · 49 models

Controls nucleus sampling by limiting generation to tokens whose cumulative probability reaches this value.

Supported by: Anthropic, DeepSeek, Mistral, OpenAI

Top K top_k integer · 22 models

Limits token sampling to the top K most likely next tokens.

Supported by: Anthropic

Frequency penalty frequency_penalty number · 13 models

Penalizes words based on how often they already appear in the generated text.

Supported by: Mistral

Presence penalty presence_penalty number · 13 models

Penalizes repeated words or phrases to encourage a wider variety of generated content.

Supported by: Mistral

Random seed random_seed integer · 13 models

Seed used for deterministic sampling when reproducible outputs are desired.

Supported by: Mistral

Seed generationConfig.seed integer · 4 models

Optional seed used for decoding when reproducible sampling is desired.

Supported by: Google

Temperature generationConfig.temperature number · 4 models

Controls randomness. Lower values make outputs more focused; higher values make them more varied.

Supported by: Google

Top K generationConfig.topK integer · 4 models

Limits token sampling to the top K most likely next tokens.

Supported by: Google

Top P generationConfig.topP number · 4 models

Controls nucleus sampling by limiting generation to tokens within the selected cumulative probability.

Supported by: Google

Reasoning

Thinking mode thinking.type enum · 21 models

Controls the Anthropic thinking mode values supported by this model.

Supported by: Anthropic, DeepSeek

Reasoning effort reasoning_effort enum · 17 models

Controls how much reasoning the model should perform before producing an answer.

Supported by: DeepSeek, OpenAI

Budget tokens thinking.budget_tokens integer · 16 models

Maximum token budget Anthropic may use for extended thinking before producing the final answer.

Supported by: Anthropic

Reasoning effort reasoning.effort enum · 9 models

Controls how much reasoning the model should perform before producing an answer.

Supported by: OpenAI

Reasoning summary reasoning.summary enum · 9 models

Controls the level of reasoning summary returned with the response.

Supported by: OpenAI

Thinking display thinking.display enum · 7 models

Controls whether Anthropic returns summarized or omitted thinking content.

Supported by: Anthropic

Include thoughts generationConfig.thinkingConfig.includeThoughts boolean · 4 models

Controls whether Gemini returns available thought summaries in the response parts.

Supported by: Google

Effort output_config.effort enum · 4 models

Controls Anthropic response thoroughness and token spend.

Supported by: Anthropic

Thinking budget generationConfig.thinkingConfig.thinkingBudget integer · 3 models

Number of thinking tokens Gemini should use; -1 uses dynamic thinking, 0 disables thinking, and fixed budgets start at 512 tokens.

Supported by: Google

Thinking level generationConfig.thinkingConfig.thinkingLevel enum · 1 model

Controls Gemini 3.5 Flash reasoning effort.

Supported by: Google

Output

Response format response_format.type enum · 13 models

Controls whether the model returns normal text or JSON mode output.

Supported by: Mistral

Verbosity text.verbosity enum · 9 models

Controls how concise or detailed the model's final text response should be.

Supported by: OpenAI

Response MIME type generationConfig.responseMimeType enum · 4 models

MIME type for generated text candidates.

Supported by: Google

Metadata

Safe prompt safe_prompt boolean · 13 models

Controls whether Mistral injects its safety prompt before the conversation.

Supported by: Mistral

Back to the full catalog

How to use

Building with an AI agent? Hit Copy to grab this whole guide as Markdown and paste it in — or point your agent straight at /llms.txt.

modelparams.dev is an open, community-maintained catalog of LLM model parameters. Each entry shows the knobs you can turn — type, default, range, and the conditions that gate it.

The same model accessed via an API key and via a subscription usually exposes a different set of parameters. We list both as separate entries so the data stays honest.

Catalog API

The full catalog is static JSON, CORS-enabled, served from the edge.

curl https://modelparams.dev/api/v1/models.json

Each entry is keyed by provider/model for API-key variants; subscription variants append -subscription.

Single model

curl https://modelparams.dev/api/v1/models/anthropic/claude-opus-4-7.json
curl https://modelparams.dev/api/v1/models/anthropic/claude-opus-4-7-subscription.json

JSON Schema

Every entry validates against a JSON Schema you can use in your editor or pipeline. 

curl https://modelparams.dev/api/v1/schema.json

Add this header to any YAML you author for autocomplete in VS Code: 

# yaml-language-server: $schema=https://modelparams.dev/api/v1/schema.json

Logos

Provider logos are available at /assets/logos/{provider}.svg where {provider} is the provider slug. They use currentColor so they inherit your text color. 

curl https://modelparams.dev/assets/logos/anthropic.svg

Logos are sourced from the models.dev repo (MIT) and used under nominative fair use.

Contribute

The data lives in YAML under models/{provider}/{model}-{auth}.yaml in the GitHub repo. Open a PR; CI validates against the schema and rebuilds.

Edit on GitHub MIT licensed