Reasoning Model Limitations (Qwen3 32b)

Last updated: May 30, 2025

This article outlines the current constraints of our only supported reasoning model: Qwen3 32b.

Key Limitations

Feature

Supported?

Notes

Disable reasoning

✅ Yes

Add /no_think in the prompt to disable reasoning

reasoning_effort

❌ No

Planned for future release)

Tool calling

✅ Yes

Supported

Structured outputs

✅ Yes

Supported

Streaming

✅ Yes

Not supported when using tool calling or SO

Usage Examples

How to disable reasoning:

{
  "model": "qwen-3-32b",
  "messages": [
    {
      "role": "system",
      "content": "/no_think\nYou are a helpful assistant."
    },
    {
      "role": "user",
      "content": "What’s the weather in New York?"
    }
  ]
}

Use /no_think in the prompt content, not as an API parameter. reasoning_effort is not available yet.

Summary

  • Qwen3 supports reasoning, tool calling, and system outputs.

  • Use /no_think in the prompt to disable reasoning.

  • reasoning_effort is not implemented yet, but will be added in future releases.

  • Streaming is incompatible with tool calling or system outputs in reasoning mode.