openclaw - 💡(How to fix) Fix [Bug]: [Bug]: xiaomi-coding/mimo-v2.5-pro returns 400 "Param Incorrect" after model switch in OpenClaw [1 comments, 2 participants]

Bug type

Regression (worked before, now fails)

Beta release blocker

No

Summary

After switching to xiaomi-coding/mimo-v2.5-pro, OpenClaw immediately returns 400 "Param Incorrect" while direct curl calls to the same API endpoint work normally.

Steps to reproduce

1.Open OpenClaw Control UI (2026.5.7 on Windows 11) 2.Switch session model to xiaomi-coding/mimo-v2.5-pro 3.Send any message (e.g. "111") 4.Observe: model fails immediately with 400 error

Expected behavior

mimo-v2.5-pro should respond normally, same as direct API calls via curl.

Actual behavior

[model-fallback/decision] decision=candidate_failed reason=format detail=400 Param Incorrect LLM request failed: provider rejected the request schema or tool payload.

OpenClaw version

2026.5.7 (eeef486)

Operating system

Windows 11

Install method

npm global

Model

xiaomi-coding/mimo-v2.5-pro

Provider / routing chain

openclaw → xiaomi-coding provider (Token Plan API: https://token-plan-cn.xiaomimimo.com/v1)

Additional provider/model setup details

"xiaomi-coding": { "baseUrl": "https://token-plan-cn.xiaomimimo.com/v1", "apiKey": "***", "api": "openai-completions", "models": [ { "id": "mimo-v2.5-pro", "name": "mimo-v2.5-pro", "reasoning": true, "compat": { "supportsTools": false } }, { "id": "mimo-v2.5", "name": "mimo-v2.5", "reasoning": true, "compat": { "supportsTools": false } } ] }

Logs, screenshots, and evidence

Gateway logs consistently show:[model-fallback/decision] decision=candidate_failed requested=xiaomi-coding/mimo-v2.5-pro candidate=xiaomi-coding/mimo-v2.5-pro reason=format next=none detail=400 Param Incorrect
Embedded agent failed before reply: LLM request failed: provider rejected the request schema or tool payload.
Direct curl test (works fine):curl -X POST https://token-plan-cn.xiaomimimo.com/v1/chat/completions \
  -H "Authorization: Bearer ******" \
  -H "Content-Type: application/json" \
  -d '{"model":"mimo-v2.5-pro","messages":[{"role":"user","content":"hi"}],"max_tokens":10}'
Returns: 200 with valid completion.

Impact and severity

Affected users: Anyone using xiaomi-coding/mimo-v2.5-pro as the active model Severity: High (completely blocks using mimo model in OpenClaw) Frequency: 100% (always fails immediately on model switch) Consequence: Must fall back to other models (minimax-portal/MiniMax-M2.7 works fine); mimo-v2.5-pro unusable through OpenClaw despite being fully functional via direct API

Additional information

Token Plan API requires reasoning_content to be preserved in multi-turn conversations with tool calls. OpenClaw may not be forwarding this field correctly to the xiaomi-coding provider, causing the API to reject the request with 400.