modelparams.dev

LLM API parameter glossary

80 API parameters appear across the catalog. This page defines each one, grouped by what it controls. Open any parameter for the full breakdown — its default, range, and conditions on every model that accepts it. Definitions come from the same community-maintained data as the JSON API.

Length

Max tokens integer max_tokens

Maximum number of output tokens the model may generate.

175 models / Alibaba, Anthropic, Cohere, DeepSeek, Fireworks, MiniMax, Mistral, NVIDIA, OpenAI, Perplexity, Thinking Machines, Z.ai

Max completion tokens integer max_completion_tokens

Maximum number of output tokens the model may generate.

100 models / Alibaba, Cerebras, Google, Groq, Meta, MiniMax, Moonshot AI, OpenAI, xAI, Xiaomi

Max tokens integer inferenceConfig.maxTokens

Maximum number of output tokens the model may generate.

56 models / Bedrock

Stop sequences string inferenceConfig.stopSequences

A list of strings where the model stops generating further tokens.

56 models / Bedrock

Stop sequence string stop

Stops generation when this string is detected.

51 models / Cerebras, Groq, Mistral, Moonshot AI, NVIDIA, xAI, Xiaomi

Max output tokens integer generationConfig.maxOutputTokens

Maximum number of tokens to include in a response candidate.

33 models / Google, Vertex

Candidate count integer generationConfig.candidateCount

How many independent completions to generate for one request.

11 models / Vertex

Stop sequences string generationConfig.stopSequences

A list of strings where the model stops generating further tokens.

11 models / Vertex

Stop sequences string stop_sequences

Stops generation when one of these sequences is detected; up to five are allowed.

8 models / Cohere

Number of completions integer n

How many chat completion choices to generate for the request.

3 models / Moonshot AI

Max output tokens integer max_output_tokens

Upper bound for output tokens generated in the Responses API response.

1 model / xAI

Sampling

Temperature number temperature

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

235 models / Alibaba, Anthropic, Cerebras, Cohere, DeepSeek, Fireworks, Google, Groq, Meta, MiniMax, Mistral, Moonshot AI, NVIDIA, OpenAI, Perplexity, Thinking Machines, xAI, Xiaomi, Z.ai

Top P number top_p

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

226 models / Alibaba, Anthropic, Cerebras, DeepSeek, Fireworks, Google, Groq, Meta, MiniMax, Mistral, Moonshot AI, NVIDIA, OpenAI, Perplexity, Thinking Machines, xAI, Xiaomi, Z.ai

Frequency penalty number frequency_penalty

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

64 models / Cerebras, Cohere, Fireworks, Groq, Meta, Mistral, Moonshot AI, NVIDIA, Xiaomi

Presence penalty number presence_penalty

Penalizes tokens that have already appeared to encourage a wider variety of content.

64 models / Cerebras, Cohere, Fireworks, Groq, Meta, Mistral, Moonshot AI, NVIDIA, Xiaomi

Top K integer top_k

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

61 models / Anthropic, Fireworks, Google, Meta

Temperature number inferenceConfig.temperature

Controls randomness. Lower values make outputs more focused; higher values make them more varied. Bedrock rejects values above 1 for this model.

56 models / Bedrock

Top P number inferenceConfig.topP

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

56 models / Bedrock

Top K integer extra_body.top_k

Limits generation to the selected number of highest-probability tokens.

41 models / Alibaba

Seed integer seed

Seed used for best-effort deterministic sampling when reproducible outputs are desired.

36 models / Cerebras, Cohere, Groq, Meta, NVIDIA, xAI

Seed integer generationConfig.seed

Optional seed used for decoding when reproducible sampling is desired.

33 models / Google, Vertex

Temperature number generationConfig.temperature

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

33 models / Google, Vertex

Top K integer generationConfig.topK

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

33 models / Google, Vertex

Top P number generationConfig.topP

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

33 models / Google, Vertex

Do sample boolean do_sample

When false, the model uses greedy decoding and ignores temperature and top_p.

22 models / Z.ai

Random seed integer random_seed

Seed used for deterministic sampling when reproducible outputs are desired.

20 models / Mistral

Top K integer additionalModelRequestFields.top_k

Limits token sampling to the top K most likely next tokens. Converse carries it through additionalModelRequestFields; it is not part of inferenceConfig.

11 models / Bedrock

Frequency penalty number generationConfig.frequencyPenalty

Penalises tokens in proportion to how often they have already appeared.

8 models / Vertex

Presence penalty number generationConfig.presencePenalty

Penalises tokens that already appeared, pushing the model toward new topics.

8 models / Vertex

Top K integer k

Limits sampling to the K most likely tokens; 0 disables top-k sampling.

8 models / Cohere

Top P number p

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

8 models / Cohere

Top K integer additionalModelRequestFields.inferenceConfig.topK

Limits token sampling to the top K most likely next tokens. Nova nests it under an inferenceConfig object inside additionalModelRequestFields.

4 models / Bedrock

Repetition penalty number repetition_penalty

Penalizes tokens that have already appeared to reduce repetition in the output.

4 models / Meta

Threshold number threshold

Confidence threshold for entity detection. Lower values detect more entities but may include false positives.

1 model / NVIDIA

Reasoning

Thinking mode enum thinking.type

Controls the Anthropic thinking mode values supported by this model.

74 models / Anthropic, Cohere, DeepSeek, Moonshot AI, Xiaomi, Z.ai

Reasoning effort enum reasoning_effort

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

52 models / Cerebras, DeepSeek, Groq, Meta, Moonshot AI, NVIDIA, OpenAI, Perplexity, Thinking Machines, xAI, Z.ai

Include thoughts boolean generationConfig.thinkingConfig.includeThoughts

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

32 models / Google, Vertex

Budget tokens integer thinking.budget_tokens

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

26 models / Anthropic

Thinking level enum generationConfig.thinkingConfig.thinkingLevel

Controls Gemini 3.1 Flash-Lite reasoning effort.

23 models / Google, Vertex

Thinking display enum thinking.display

Controls whether Anthropic returns summarized or omitted thinking content.

22 models / Anthropic

Enable thinking boolean extra_body.enable_thinking

Toggles the model's hybrid thinking mode, sent as a provider-specific extra body field on OpenAI-compatible clients.

19 models / Alibaba

Thinking budget integer extra_body.thinking_budget

Maximum number of tokens the model may spend on reasoning before it starts the final answer; defaults to the model's maximum reasoning length.

19 models / Alibaba

Thinking budget integer generationConfig.thinkingConfig.thinkingBudget

Token budget the model may spend on internal reasoning before answering. 0 disables thinking. Cannot be combined with thinkingLevel.

19 models / Google, Vertex

Enable thinking boolean extra_body.chat_template_kwargs.enable_thinking

Controls Qwen3 thinking mode when using OpenAI-compatible clients that pass provider-specific extra body fields.

18 models / Alibaba

Effort enum output_config.effort

Controls Anthropic response thoroughness and token spend.

16 models / Anthropic

Reasoning effort enum reasoning.effort

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

15 models / OpenAI, xAI

Reasoning summary enum reasoning.summary

Controls the level of reasoning summary returned with the response.

14 models / OpenAI

Split reasoning boolean reasoning_split

Returns the model's reasoning in a separate reasoning_details field instead of inline with the response.

8 models / MiniMax

Budget tokens integer additionalModelRequestFields.thinking.budget_tokens

Maximum token budget the model may use for extended thinking before producing the final answer. Must be less than inferenceConfig.maxTokens.

7 models / Bedrock

Thinking mode enum additionalModelRequestFields.thinking.type

Controls Anthropic extended thinking on Bedrock. Bedrock rejects the adaptive mode the first-party Anthropic API accepts, so only disabled and enabled are listed here.

7 models / Bedrock

Reasoning format enum reasoning_format

Controls how reasoning tokens are returned — hidden from the response, raw within the content, or parsed into a separate field.

5 models / Groq

Reasoning budget integer reasoning_budget

Maximum number of tokens the model may use for internal reasoning before being forced to end the reasoning trace. Use -1 to disable budget enforcement.

3 models / NVIDIA

Prompt mode enum prompt_mode

Enables Mistral's reasoning system prompt; leave unset to disable the default reasoning behavior.

2 models / Mistral

Keep prior reasoning enum thinking.keep

Set to all to carry reasoning content from earlier assistant turns into the request; leave null to send only the final answers.

2 models / Moonshot AI

Clear thinking boolean clear_thinking

When true, the model's thinking from previous turns is excluded from the conversation context; when false, it is preserved, which is useful for agentic workflows.

1 model / Cerebras

Separate reasoning boolean separate_reasoning

Returns the model's reasoning in a dedicated reasoning_content field instead of interleaving it with the final message content.

1 model / Thinking Machines

Thinking token budget integer thinking.token_budget

Maximum number of tokens the model may spend on reasoning before answering.

1 model / Cohere

Tools

Tool choice enum tool_choice

Controls whether the model may call tools, must call one, or skips tool calls.

21 models / Cohere, Meta, Moonshot AI, OpenAI

Output

Response format enum response_format.type

Forces the response into plain text or a JSON object.

106 models / Cerebras, Cohere, Fireworks, Groq, Meta, Mistral, Moonshot AI, OpenAI, xAI, Xiaomi, Z.ai

Response MIME type enum generationConfig.responseMimeType

MIME type for generated text candidates.

33 models / Google, Vertex

Verbosity enum text.verbosity

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

14 models / OpenAI

Text format enum text.format.type

Controls whether the Responses API returns free-form text, JSON mode output, or structured JSON schema output.

1 model / xAI

Observability

Log probabilities boolean logprobs

Controls whether the response includes log probabilities for the generated tokens.

15 models / Cohere, Moonshot AI

Top log probabilities integer top_logprobs

Number of most likely tokens to return a log probability for at each position.

7 models / Moonshot AI

Metadata

Safe prompt boolean safe_prompt

Controls whether Mistral injects its safety prompt before the conversation.

20 models / Mistral

Safety mode enum safety_mode

Controls Cohere's built-in safety instructions applied to the generation.

8 models / Cohere

Return images boolean return_images

Controls whether the response may include related images from the search.

4 models / Perplexity

Return related questions boolean return_related_questions

Controls whether the response includes suggested follow-up questions.

4 models / Perplexity

Search after date string search_after_date_filter

Restricts search results to content published after this date (MM/DD/YYYY).

4 models / Perplexity

Search before date string search_before_date_filter

Restricts search results to content published before this date (MM/DD/YYYY).

4 models / Perplexity

Search domain filter string search_domain_filter

Limits search to, or excludes, specific domains.

4 models / Perplexity

Search mode enum search_mode

Selects the corpus the model searches when grounding its answer.

4 models / Perplexity

Search recency filter enum search_recency_filter

Restricts web search results to a recent time window.

4 models / Perplexity

Search context size enum web_search_options.search_context_size

Controls how much web search context is retrieved before generating the answer.

4 models / Perplexity

Disable search boolean disable_search

Turns off web search so the model answers from its own knowledge only.

3 models / Perplexity

Prompt cache retention enum prompt_cache_retention

Controls whether the prompt cache stays in memory or persists for up to 24 hours.

2 models / Meta

Chunk length integer chunk_length

Context window size for processing. Longer texts are automatically split into chunks with overlap for complete coverage. Must be greater than overlap.

1 model / NVIDIA

Expert type enum expert_type

The type of expert to use. 'knowledge' answers with USD knowledge, 'code' responds with vanilla OpenUSD code, 'helperfunction' uses high-level helper functions, and 'auto' lets the LLM determine which expert to use.

1 model / NVIDIA

Flat NER boolean flat_ner

When true, prevents overlapping entity spans. When false, may return nested entities such as both a full name and its constituent first name.

1 model / NVIDIA

Overlap integer overlap

Token overlap between chunks to prevent entity clipping. Must be less than chunk_length.

1 model / NVIDIA

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 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.

If you only need the params for one model contract, use the providerless endpoint. Subscription contracts are model slugs with -subscription.

curl https://modelparams.dev/api/v1/models/openai/gpt-5.5.json
curl https://modelparams.dev/api/v1/models/openai/gpt-5.5-subscription.json

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