Bug type

Behavior bug (incorrect output/state without crash)

Beta release blocker

No

Summary

connecting 2 local machines with openclaw and ollama and using a small model. This is a test scenario.

Steps to reproduce

OpenClaw 2026.4.12 on 192.168.1.101, Ollama on 192.168.1.102:11434.

openclaw infer model run --model ollama/qwen2.5:0.5b --prompt "Reply with exactly: hello" --json
fails with:
400 {"error":"\"qwen2.5:0.5b\" does not support thinking"}

But direct Ollama calls to /api/chat with the same model work, including with "think": false.

Expected behavior

The agent does not respond as this is the first time connecting these 2 devices together.

Actual behavior

no response in the openclaw gateway ui.

OpenClaw version

OpenClaw 2026.4.12

Operating system

Linux 24:04

Install method

No response

Model

ollama/qwen2.5:0.5b but changing to another model

Provider / routing chain

openclaw local 192,168.1.101 to ollama 192.168.1.102

Additional provider/model setup details

No response

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

No response