Parameters by Upstream
Some request fields do not mean the same thing everywhere. This page states what the gateway does with each one on the way to each provider - passed through, translated into the provider's own equivalent, or dropped.
This is not a list of every accepted field; each endpoint page documents its own request body. Listed here are the fields where supported is not a yes or no, because the answer changes with the provider your model runs on.
reasoning_effort
Full guide, with values and cost impact →
| You call | Upstream | Behaviour | Descripción |
|---|---|---|---|
/v1/chat/completions | OpenAI-compatible (GPT, GLM, DeepSeek, Qwen, Kimi…) | transmitido | Forwarded to the provider as sent, with one exception: when a model's catalog entry lists the effort levels it accepts, a value below the lowest is raised to it and one above the highest steps down (GPT-6.1 Sol cannot turn reasoning off, so none and minimal run at low). Otherwise the gateway neither validates the value nor guarantees the provider honours it. Commonly minimal, low, medium and high, but the accepted set is the provider's: OpenAI also takes none on most models, GLM takes up to max. Check the provider's own reference for its range. |
/v1/chat/completions | Google Vertex - Gemini 4 | traducido | Becomes Gemini's thinkingLevel, kept inside the three levels Gemini 4 Argon accepts: none and minimal run at low (the model cannot turn thinking off), low/medium/high map straight across, xhigh and max run at high. An explicit google.thinking_config in the request wins over this mapping. |
/v1/chat/completions | Google Vertex - Gemini 3.x | traducido | Becomes Gemini's thinkingLevel: none and minimal both map to minimal, low/medium/high map straight across. An explicit google.thinking_config in the request wins over this mapping. |
/v1/chat/completions | Google Vertex - Gemini 2.5 | descartado | Gemini 2.5 only understands thinkingBudget and rejects thinkingLevel outright, so the gateway sends neither and leaves the model at its default thinking behaviour. Use google.thinking_config with thinking_budget to control 2.5. |
/v1/chat/completions | Anthropic (Claude) | traducido | Se convierte en el thinking de Anthropic, en la forma que acepta el modelo. En los modelos Claude que solo aceptan razonamiento adaptativo (Claude Opus 4.7 y posteriores, Sonnet 5 y posteriores, Haiku 5.5), low/medium/high/xhigh/max pasan a thinking:{"type":"adaptive"} con output_config.effort al mismo nivel, y minimal se ejecuta como low. none desactiva el razonamiento en la forma propia del modelo: {"type":"disabled"} en la mayoría, {"type":"between_tools"} en Sonnet 5.5, y adaptativo en low en Opus 5.5, que no puede desactivar el razonamiento. Cualquier otro valor no se reenvía. En los modelos Claude anteriores se convierte en un presupuesto: none y minimal desactivan el razonamiento ({"type":"disabled"}); low/medium/high/xhigh/max pasan a un presupuesto de 2,048 / 8,192 / 16,384 / 32,768 tokens, la inversa del reparto usado en la otra dirección. Un presupuesto derivado se mantiene por debajo de max_tokens y no se envía en absoluto cuando max_tokens no alcanza el mínimo de 1,024 tokens de Anthropic, así que pedir un nivel de esfuerzo nunca convierte una petición que funciona en un error 400 del proveedor. |
/v1/messages | OpenAI-compatible (GPT, GLM, DeepSeek, Qwen, Kimi…) | traducido | Derived from thinking.budget_tokens by bucket, when extended thinking is enabled and the budget is above zero: up to 2,048 tokens becomes low, up to 8,192 becomes medium, above that becomes high. A zero or absent budget sends nothing. For a model whose catalog entry lists the effort levels it accepts (GPT-6.1 Sol), the request's intent is followed instead: thinking disabled or between_tools runs at the lowest level (low), even when output_config.effort is also set; otherwise output_config.effort is sent as the effort. |
/v1/responses | OpenAI-compatible (GPT, GLM, DeepSeek, Qwen, Kimi…) | traducido | Read out of the Responses-API reasoning.effort object and sent as the flat reasoning_effort field. |
thinking / thinking_budget / enable_thinking
| You call | Upstream | Behaviour | Descripción |
|---|---|---|---|
/v1/chat/completions | Anthropic (Claude) | traducido | Son los nombres que otros proveedores dan al mismo control, y la pasarela los normaliza. thinking:{"type":"disabled"}, enable_thinking:false y thinking_budget:0 desactivan el razonamiento, igual que reasoning_effort none en ese modelo. En los modelos Claude anteriores, thinking_budget:N y thinking:{"type":"enabled","budget_tokens":N} reenvían el presupuesto sin cambios; un valor fuera de rango lo rechaza el proveedor en lugar de ajustarse en silencio. En los modelos que solo aceptan razonamiento adaptativo nunca se envía un presupuesto: se convierte en output_config.effort (hasta 2,048 low, hasta 8,192 medium, por debajo de 32,768 high, desde 32,768 xhigh; nunca max). thinking:{"type":"adaptive"} por sí solo no envía ningún ajuste de razonamiento, y el modelo se ejecuta con su comportamiento predeterminado. |
/v1/chat/completions | Google Vertex - Gemini 4 | traducido | Normalised into thinkingConfig.thinkingLevel, because Gemini 4 Argon rejects thinkingBudget: turning thinking off runs at low, and an explicit budget is folded into a level (up to 2,048 tokens low, up to 8,192 medium, above that high). An explicit google.thinking_config in the request still wins. |
/v1/chat/completions | Google Vertex - Gemini 3.x | traducido | Normalised into thinkingConfig: turning thinking off becomes thinkingLevel:"minimal", and an explicit budget becomes thinkingBudget. An explicit google.thinking_config in the request still wins over all of it. |
/v1/chat/completions | Google Vertex - Gemini 2.5 | traducido | Normalised into thinkingConfig.thinkingBudget - turning thinking off sends a budget of 0, and an explicit budget passes through. thinkingLevel is not used here because 2.5 rejects it. An explicit google.thinking_config in the request still wins. |
Every row is transcribed from the converter that implements it and pinned by a Go test, so this table cannot drift from the running gateway without a build failing.