GLM 4.5 OR Chutes AI error 400

[markp018]

App Version

3.25.8

API Provider

OpenRouter

Model Used

Z.al/GLM 4.5

Roo Code Task Links (Optional)

No response

🔁 Steps to Reproduce

Just asking any question.

💥 Outcome Summary

Error:

{
"object": "error",
"message": "Requested token count exceeds the model's maximum context length of 98304 tokens. You requested a total of 115623 tokens: 10527 tokens from the input messages and 105096 tokens for the completion. Please reduce the number of tokens in the input messages or the completion to fit within the limit.",
"type": "BadRequestError",
"param": null,
"code": 400
}

When I switched back to v3.25.7, the problem is gone.

📄 Relevant Logs or Errors (Optional)

[roomote]

I'm here to help out the maintainers and am going to investigate this issue with GLM 4.5 on OpenRouter exceeding the token limit. I'll look into what changed between v3.25.7 and v3.25.8 that could be causing the model to request more tokens than its maximum context length.

[roomote]

I've successfully created a fix for this issue! 🎉

The problem was that commit c52fdc4 introduced a 20% clamping threshold for model max tokens, which was too restrictive for GLM-4.5 models that legitimately need 75% of their context window for output tokens.

Solution implemented in PR #6807: #6807

The fix adjusts the clamping threshold from 20% to 80%, which:

• Still prevents models from using the entire context for output
• Allows models like GLM-4.5 to function properly with their required 98,304 output tokens
• Only applies clamping when truly necessary

All tests are passing and CI checks have completed successfully. The PR is ready for review!