[gpt-5.*] 'max_tokens' is not supported with this model, when using azure BYOK

[zaherg]

When routing to Azure via BYOK, OpenRouter forwards max_tokens as-is instead of translating it to max_completion_tokens.

Since GPT-5* endpoint has deprecated max_tokens, the call will fall

{
  "error": {
    "message": "Unsupported parameter: 'max_tokens' is not supported with this model. Use 'max_completion_tokens' instead.",
    "type": "invalid_request_error",
    "param": "max_tokens",
    "code": "unsupported_parameter"
  }
}

Response metadata:

{
  "provider_name": "Azure",
  "is_byok": true
}

Setup:

• Model: openai/gpt-5.2-chat
• Provider routing: only: ["azure"]
• Using @openrouter/ai-sdk-provider@2.1.1 and ai@6.0.49 (Vercel AI SDK)
• No max_tokens or max_completion_tokens explicitly set in the request

Expected behavior:
OpenRouter should translate max_tokens to max_completion_tokens when forwarding to Azure for GPT-5 models, as it does for non-BYOK requests.

Workaround:
Routing to a non-Azure BYOK provider avoids the issue.