# Parameters Every request field of chat completions — what the router checks, what it changes, what it passes to the provider and where each field works. ## How the router treats parameters 1. **Checked by the router:** `model`, `models`, `messages`, `stream`, `max_tokens`, `max_completion_tokens`, `tools`, `response_format` and `provider`. A wrong type or value is a `400` that names the field in `error.metadata.fields`. 2. **Passed through:** any other field goes to the provider unchanged. Whether the model supports it depends on the model: see `supported_parameters` of the model in `GET /api/v1/models` or in the [catalog](/models). A model ignores parameters it does not support. 3. **Dropped in the Russia region:** for `ru-1/…` and `ru-2/…` models the router silently removes `provider`, `models`, `route`, `transforms`, `plugins`, `usage`, `include_reasoning`, `reasoning`, `web_search_options`, `prediction` and `debug`. `reasoning.effort` becomes `reasoning_effort`. > **Warning:** Without `max_tokens` or `max_completion_tokens` the router sets `max_tokens` to 4096: a longer answer ends with `finish_reason: "length"`. Set it explicitly for long answers and for reasoning models, whose reasoning tokens count toward the limit. In the cards below, **Regions** says where a field works: Global (`vendor/model` ids) and Russia (`ru-1/…`, `ru-2/…`). ## Model and messages ### `model` Type: `string` · Regions: Global, Russia The model id from the catalog. Required unless you give `models`. ### `models` Type: `string[]` · Regions: Global, Russia Fallback models tried in order after `model`, if a model has no route or fails before the first token. At most 3 attempts per request in total. See [Models and routing](/docs/en/models-and-routing#fallback-models). ### `messages` Type: `object[]` · Required · Regions: Global, Russia The conversation. Every message needs a `role`. See [Chat completions](/docs/en/chat-completions#messages-and-roles). ### `stream` Type: `boolean` · Default: `false` · Regions: Global, Russia Send the answer as server-sent events. See [Streaming](/docs/en/streaming). ## Output length ### `max_tokens` Type: `integer` · Default: `4096` · Range: 1 to 65536 · Regions: Global, Russia The most tokens the model may generate. Before the request the router reserves funds for this many tokens (at most the model's own output limit), so a smaller value also needs a smaller balance. Values above 65536 are rejected with `400`. ### `max_completion_tokens` Type: `integer` · Range: 1 to 65536 · Regions: Global, Russia The newer name of the same limit. When both are set, the smaller one applies to both. ## Sampling ### `temperature` Type: `number` · Range: 0 to 2 · Regions: passed through unchecked Randomness of the output: lower is more predictable, higher is more varied. The default depends on the model, usually 1. ### `top_p` Type: `number` · Range: 0 to 1 · Regions: passed through unchecked Nucleus sampling: only the tokens within this cumulative probability are considered. Change either `temperature` or `top_p`, not both. ### `top_k` Type: `integer` · Regions: passed through unchecked Only the K most likely tokens are considered. `0` turns it off. Not every model supports it. ### `frequency_penalty` Type: `number` · Range: -2 to 2 · Regions: passed through unchecked Penalizes tokens by how often they already appeared; positive values reduce repetition. ### `presence_penalty` Type: `number` · Range: -2 to 2 · Regions: passed through unchecked Penalizes any token that already appeared; positive values push the model to new topics. ### `repetition_penalty` Type: `number` · Range: 0 to 2 · Regions: passed through unchecked Reduces repetition of tokens from the input; `1` means no penalty. Mostly for open-weight models. ### `min_p` Type: `number` · Range: 0 to 1 · Regions: passed through unchecked The smallest probability of a token relative to the most likely one. ### `seed` Type: `integer` · Regions: passed through unchecked Makes sampling repeatable where the provider supports it. The same seed and parameters usually, not always, give the same answer. ### `stop` Type: `string | string[]` · Regions: passed through unchecked Sequences that end generation; the sequence itself is not returned. ### `logit_bias` Type: `object` · Regions: passed through unchecked Token ids mapped to a bias from -100 to 100 that makes those tokens more or less likely. ### `logprobs` Type: `boolean` · Regions: passed through unchecked Return the log probabilities of the output tokens. `top_logprobs` (0 to 20) also returns the most likely alternatives. ## Tools and structured output ### `tools` Type: `object[]` · Regions: Global, Russia Functions the model may call, in the OpenAI format. A non-empty list sends the request only to routes that support tools. See [Tool calling](/docs/en/tool-calling). ### `tool_choice` Type: `string | object` · Regions: passed through unchecked `auto`, `none`, `required`, or an object that names one function. ### `parallel_tool_calls` Type: `boolean` · Regions: passed through unchecked Whether the model may request several tool calls in one answer. ### `response_format` Type: `object` · Regions: Global, Russia `{"type": "json_object"}` asks for valid JSON; `{"type": "json_schema", "json_schema": …}` asks for JSON that matches a schema and sends the request only to routes that support structured output. See [Structured outputs](/docs/en/structured-outputs). ## Reasoning ### `reasoning` Type: `object` · Regions: Global Controls the thinking of reasoning models: `effort` (`minimal`, `low`, `medium`, `high`), or `max_tokens` for the reasoning budget, `exclude: true` to leave the reasoning out of the answer, `enabled`. In the Russia region only `effort` is kept, as `reasoning_effort`. See [Reasoning](/docs/en/reasoning). ### `reasoning_effort` Type: `string` · Regions: passed through unchecked The OpenAI shortcut for `reasoning.effort`. ## Routing and privacy ### `provider` Type: `object` · Regions: Global, Russia Where the request may go. The router itself applies: - `only`, `ignore`, `order`: lists of provider slugs; the slug `ru` names the Russia region; - `allow_fallbacks` (boolean): `false` keeps one route per model; - `zdr` (boolean): `true` sends the request only to models with a zero data retention endpoint; - `data_collection` (`allow` or `deny`): `deny` excludes providers that may store or train on data. Other keys of the object go to the Global region as is, without guarantees. See [Models and routing](/docs/en/models-and-routing) and [Zero data retention](/docs/en/zero-data-retention). ## Other fields ### `user` Type: `string` · Regions: passed through unchecked Your id of the end user. The router does not store it; some providers use it to detect abuse. ### `usage` Type: `object` · Regions: Global Ignored: the router always asks for usage, and every response has `usage` with `cost`. ### `stream_options` Type: `object` · Regions: passed through unchecked `include_usage` is always on for streams: the last chunk before `[DONE]` carries `usage`. Fields that are not listed here are passed to the provider as they are. The router never returns an error for an unknown field, but the model may ignore it.