Skip to content
CC Switch WikiUnofficial Community Wiki

Model always thinks; disable not supported; use low, high, or max. - CC Switch

Quick fix

Set glm-5.3 to forced thinking in cc-switch config; disable thinking.type: disabled.

Symptom

Section titled “Symptom”
Error output
该模型始终思考，不支持关闭思考；请使用 low、high 或 max。
English translation

The model always thinks; disabling thinking is not supported; please use low, high, or max.

Cause

Section titled “Cause”

When routing requests from Claude Code to Zhipu GLM-5.3 (or GLM-5.3-Flash) via cc-switch, the upstream client sends `thinking.type: disabled` to turn off the thinking process. However, the GLM-5.3 series models enforce thinking and do not support the `disabled` state. The cc-switch proxy layer forwards this parameter by default, causing the upstream API to return an HTTP 400 error with business code 1210.

Additionally, GLM-5.3 strictly limits reasoning effort to only `low`, `high`, and `max`. Requests carrying illegal values or failing to specify the correct enabled state will also fail.

Fix

Section titled “Fix”

  1. In cc-switch supplier configuration or routing rules, locate settings for glm-5.3 or glm-5.3-flash.

  2. Modify thinking control logic: change `thinking.type` from `disabled` to `enabled`, and set a reasonable `reasoning_effort` (e.g., `low`).

    伪代码示例：在 proxy 转换层或 provider 配置中
    # 原配置/行为:
    "thinking": { "type": "disabled" }
    # 修正为:
    "thinking": { "type": "enabled", "reasoning_effort": "low" }

  3. Restart cc-switch proxy service to ensure new routing logic takes effect.

Affected Versions

Section titled “Affected Versions”
ToolClaude Code
Version3.20.0
PlatformsmacOSWindows

Source Issues

Section titled “Source Issues”

This page is distilled from 2 real issues

FAQ

Section titled “FAQ”
Why do other GLM models (e.g., glm-5.2) not have this issue?
GLM-5.2 and earlier versions support switching thinking modes (including disabling), whereas the GLM-5.3 architecture changes enforce thinking, making it incompatible with the `disabled` parameter.
Will I still encounter this error if I do not use GLM-5.3?
No. This issue is specific to GLM-5.3 and GLM-5.3-Flash models. Other models are handled according to original logic.
Section titled “Related problems”
usage null - Claude Code Qwen DashScopeUpgrade to cc-switch v3.16.3+, which fixes null streaming usage on the OpenAI-compatible interface automatically.2 issues mergedapi failed deserialize json body intoDepending on the error: roll Claude Code back, upgrade cc-switch, turn the local proxy off, or stop sending unsupported fields.7 issues mergedRange of input length should be [1, 1048576]Upgrade CC Switch to v3.19.0 for the fix where images were forwarded as base64 text and inflated tokens, or avoid view_image under the proxy.3 issues merged

This is an unofficial community wiki with no affiliation to the cc-switch authors or the project itself. Its content is compiled from the project's public GitHub issues. This site distributes no software.