#Using more than 3 "models" returns 400 and possible issue with gemini-2.5-flash-preview.

15 messages · Page 1 of 1 (latest)

tired fiber
#

Issue 1: Attempting to have more than 3 models in the models parameter results in a 400.

The identical request removing any model succeeds. If this is a planned limitation, it should really be documented.

Requests payload (formatted for readability)

{
  "model": "google/gemini-2.5-flash-preview",
  "models": [
    "google/gemini-2.0-flash-001",
    "qwen/qwq-32b",
    "openai/gpt-4.1-mini",
    "deepseek/deepseek-chat-v3-0324"
  ],
  "messages": [
    {
      "role": "system",
      "content": "You are a multi-lingual translator. The user may provide text in any language to be translated.\nReturn translations for these languages: **[\"es\"]**.\n\n• Your goal is to **produce translations** that accurately reflect the **original meaning**, while maintaining an overall **friendly** tone.\n• The current user is: **doc**.\n• The entire chat history (if any) will be provided as context, but **do not translate the history itself**.\n• **Use context selectively** to clarify or enhance meaning **without changing the intent**. For instance, if a message references a choice made earlier, you can include that in the translation if it improves clarity—but **do not invent or assume** details not given.\n• **Do not assume** every message is a direct reply to the previous one. Consider each message on its own unless there is **clear contextual evidence** they are connected.\n• **Match the original tone and level of formality** as closely as possible. If the tone is ambiguous, choose a **friendly** style in line with casual conversation, but **avoid adding extra flair or significantly altering** the message.\n• **Avoid drastically changing or expanding** words, phrases, or greetings (e.g., do not change \"Hi\" to \"Howdy\").\n\nOutput must be valid JSON strictly matching the provided schema."
    },
    {
      "role": "user",
      "content": "Translate this message from doc:\n\ntest message"
    }
  ],
  "provider": {
    "require_parameters": true
  },
  "response_format": {
    "type": "json_schema",
    "json_schema": {
      "name": "multi_translation",
      "strict": true,
      "schema": {
        "type": "object",
        "properties": {
          "translations": {
            "type": "object",
            "properties": {
              "es": {
                "type": "string",
                "description": "Translation in language es."
              }
            },
            "required": ["es"],
            "additionalProperties": false
          }
        },
        "required": ["translations"],
        "additionalProperties": false
      }
    }
  },
  "temperature": 0.5,
  "max_tokens": 1000,
  "top_p": 1,
  "frequency_penalty": 0,
  "presence_penalty": 0
}

Issue 2 - google/gemini-2.5-flash-preview returns 404 with structured outputs

This same request (without the models parameter) was working fine until about 12 hours ago when I started getting a 404. I believe it is due to the structured outputs, however the documentation on the web lists that structured outputs and request format should be supported. Not sure what's going on there.

Obviously with the models parameter (provided there are no more than 3) this works but it is using a fallback model.

tough falcon
#

What’s the 400 error message?

tired fiber
#

[2025-05-02 23:50:54] [WARNING ] autoTranslate: Error on attempt 1 translating text: 400 Client Error: Bad Request for url: https://openrouter.ai/api/v1/chat/completions

tough falcon
#

That looks like an intermediate later

#

Not the body of the 400 response

tired fiber
#

1 sec and I can give you the full output.

#

well how about that... the full response actually says its limited to 3 models. Would be very nice if that was anywhere in the documentation.

#

[2025-05-03 00:02:30] [DEBUG ] autoTranslate: Json: {'error': {'message': "'models' array must have 3 items or fewer.", 'code': 400}, 'user_id': 'xxx'}

#

Issue 2 stands however
[2025-05-03 00:07:24] [DEBUG ] autoTranslate: Json: {'error': {'message': 'No endpoints found that can handle the requested parameters. To learn more about provider routing, visit: https://openrouter.ai/docs/provider-routing', 'code': 404}}

tired fiber
#

Bump? Or am I missing something here?

tough falcon
#

Answer is on the model page

image0.jpg
#

You passed require_parameters: true

#

Along with frequency and presence penalty. Delete require_parameters: true in your code

tired fiber
#

Missed that on my config. Interesting that that changed the other day.

I would highly recommend that the documentation is updated to reflect the 3 model max in the models param.

Cheers.

tough falcon
#

Ah I think it changed because we learned that Flash 2.5 does not support frequency and presence penalty. Sorry about that! Will think more about how to make this clear