Bug type

Behavior bug (incorrect output/state without crash)

Beta release blocker

No

Summary

In OpenClaw 2026.6.8, ollama/glm-5.2:cloud is discovered from the local Ollama endpoint and shown by openclaw models list, but Telegram /think still lists only default/off while /think medium is accepted for the same session/model.

Steps to reproduce

Shortest observed repro:

openclaw --version
# OpenClaw 2026.6.8 (844f405)

openclaw config unset models.providers.ollama.models
openclaw config set agents.defaults.models '{"ollama/*":{}}' --strict-json --replace
openclaw gateway restart

curl -s http://127.0.0.1:11434/api/tags | jq -r '.models[].name' | grep '^glm-5\.2:cloud$'
curl -s http://127.0.0.1:11434/api/show -H 'content-type: application/json' -d '{"name":"glm-5.2:cloud"}'
openclaw models list --provider ollama

Then in Telegram:

/model ollama/glm-5.2:cloud
/think
/think medium

Observed result: /think shows only default, off, but /think medium returns Thinking level set to medium.

Expected behavior

Because the same endpoint reports glm-5.2:cloud capabilities as ["thinking", "completion", "tools"], and /think medium is accepted for the selected session model, bare /think should show the supported Ollama thinking levels for that model, e.g. default, off, low, medium, high, max, rather than only default, off.

Actual behavior

openclaw --version reports OpenClaw 2026.6.8 (844f405). curl /api/tags returns glm-5.2:cloud, curl /api/show for that model returns capabilities: ["thinking", "completion", "tools"], and openclaw models list --provider ollama shows ollama/glm-5.2:cloud with text input and 977k context. In Telegram, after /model ollama/glm-5.2:cloud, bare /think reports Current thinking level: high but lists only Options: default, off; sending /think medium then returns Thinking level set to medium.

OpenClaw version

2026.6.8

Operating system

Debian 13

Install method

nom global

Model

glm-5.2

Provider / routing chain

openclaw -> local Ollama Docker instance -> Ollama Cloud -> glm-5.2

Additional provider/model setup details

No response

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

Related issues/PRs that are likely related to this since I ran into a very similar issue when MiniMax M3 dropped on Ollama Cloud:
#90315
#91248