GLM-4.5 is not supported. Unable to generate a response after thinking.

[pilgrim1385]

What happened?

GLM-4.5 is not supported. Unable to generate a response after thinking.

Version Information

ghcr.io/danny-avila/librechat-dev latest b2acb22fc0ee 10 hours ago 1.16GB
ghcr.io/danny-avila/librechat-rag-api-dev-lite latest 833a22bcf14c 2 weeks ago 1.45GB

Steps to Reproduce

1. Config in librechat.yaml

    - name: "ZhiPu"
      apiKey: "${ZHIPU_API_KEY}"
      baseURL: "https://open.bigmodel.cn/api/paas/v4"
      models:
        default: ["glm-4.5", "glm-4.5-air", "glm-4.5-x", "glm-4.5-airx", "glm-4.5-flash"]
        fetch: true
      titleConvo: false
      titleModel: ""
      summarize: false
      summaryModel: ""
      forcePrompt: false
      modelDisplayLabel: "ZhiPu"

2. Run docker compose up -d
3. Chat in LibreChat

What browsers are you seeing the problem on?

Chrome

Relevant log output

N/A

Screenshots

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[onestardao]

This sounds like one of those cases where the API call goes through, but LibreChat doesn't know how to handle the response format — especially if the model family (like GLM) isn't using standard OpenAI-compatible returns or needs a custom prompt template.

Couple of things worth checking:

• Does the GLM-4.5 API require a special system prompt or prefix format?
• Is the model response wrapped in a field LibreChat doesn’t expect (e.g., response.content vs text)?
• Have you tried testing that model separately via curl or Postman to confirm the API is responding correctly?

I’ve seen a few cases like this — where the model is actually responding, but the frontend just gets stuck on “thinking…” because the parser fails silently.

Let me know if you're able to confirm the raw response from the GLM API. Could help isolate whether it’s a frontend template mismatch or an actual backend failure.

[danny-avila]

Looks like some models work better than others, no issues with glm-4.5-air:

glm-4.5 response stream gets cut short after thinking and the issue is not from LC.

[danny-avila]

I'm double checking with this API to confirm

[danny-avila]

Closed by #8769

[Odrec]

I updated to the latest version and can confirm it is using the updated agents dependency but glm-4.5 still is not working. No web search.

also if not using the tool then the response goes weirdly above the thinking area

[Odrec]

Just to add we're self-hosting GLM-4.5 full version with mlx-lm.