ollama - 💡(How to fix) Fix Bug: qwen3.5:397b-cloud hangs on /v1/chat/completions endpoint [1 comments, 2 participants]

Model

• qwen3.5:397b-cloud

Description

The model qwen3.5:397b-cloud hangs on the /v1/chat/completions endpoint while /api/generate works fine.

Steps to Reproduce

1. Start Ollama with qwen3.5:397b-cloud model loaded
2. Send POST request to /v1/chat/completions
3. Request hangs/times out with 500 Internal Server Error

Expected Behavior

Chat completion should return within reasonable time

Actual Behavior

• /api/generate works fine
• /v1/chat/completions hangs/times out (500 Internal Server Error)
• Other models (qwen3.5:cloud, minimax-m2.5:cloud) work fine on same server with both endpoints

Investigation

# Works:
curl -X POST http://localhost:11434/api/generate -d '{"model": "qwen3.5:397b-cloud", "prompt": "Hi"}'

# Hangs/Timeouts:
curl -X POST http://localhost:11434/v1/chat/completions -d '{"model": "qwen3.5:397b-cloud", "messages":[{"role":"user","content":"Hi"}]}'

Notes

• Model size: 397b parameters
• Same model works on /api/generate but not on /v1/chat/completions
• Other models work on both endpoints without issues

Created via GitHub API