Claude Code always gets API error in the new release, v1.83.10. API Error: 400 {"error":{"message":"{\"type\":\"error\",\"error\":{\"type\":\"invalid_request_error\",\"message\":\"This model API Error: 400 {"error":{"message":"{"type":"error","error":{"type":"invalid_request_error","message":"adaptive After an investigation, it seems that for supporting opus-4.7, the newly introduced function, _translate_legacy_thinking_for_adaptive_model, in litellm/llms/anthropic/experimental_pass_through/messages/transformation.py, always sets effort level to xhigh which results in invalid request error. 3. API Error: 400 returned Error Code: 400 Message: {"type":"error","error":{"type":"invalid_request_error","message":"This model does not support effort level 'xhigh'. Supported levels: high, low, max, medium."},"request_id":"req_011CahfRsaJuspamBxb1zcyb"}

TL;DR

Set the effort level to a supported value, such as "high", in the output_config of litellm_params to mitigate the API error.

Guidance

• The issue is caused by the _translate_legacy_thinking_for_adaptive_model function setting the effort level to "xhigh", which is not supported by the model.
• To verify the fix, launch Claude Code with the modified litellm_params and check for API errors.
• As a temporary workaround, set "output_config": { "effort": "high" } in litellm_params to avoid the invalid request error.
• Review the transformation.py file to ensure that the effort level is set correctly for different models.

Example

litellm_params = {
    "output_config": {
        "effort": "high"
    }
}

Notes

This fix assumes that the issue is solely caused by the unsupported effort level. If other errors occur, further investigation may be needed.

Recommendation

Apply workaround: set the effort level to a supported value, such as "high", in the output_config of litellm_params. This is because the newly introduced function is causing the issue, and modifying the litellm_params provides a quick mitigation.