https://models.ageneral.ai uses OpenAI-compatible POST /v1/chat/completions and POST /v1/responses bodies. Your JSON is forwarded to inference as you send it. Ageneral may apply product defaults, such as filling model when you omit it.
That means standard fields such as temperature, top_p, max_tokens or max_completion_tokens, stop, tools, tool_choice, response_format, stream, and others in the same family are accepted on the wire the same way as on other OpenAI-compatible gateways.
What actually applies
Whether a given parameter changes behavior depends on the provider and model behind the route, not on Ageneral stripping fields. Some models ignore or partially support certain options; others match OpenAI’s or ecosystem docs closely. If something has no effect, treat it as a model or provider limitation, not a missing feature on the Models host.References
| Topic | Link |
|---|---|
| Chat Completions fields | OpenAI — Create chat completion |
| Responses fields | OpenAI — Create response |
| Common optional and sampling parameters — OpenRouter catalog | OpenRouter — API parameters |