Can you turn off thinking on GLM-5.3?
No. Both GLM models refuse every documented way of disabling it, and their effort levels are genuinely graded.
What we found
Both GLM models served here are thinking-only. There are two documented ways to ask a model to stop reasoning, and we tried both against the live API. Sending reasoning_effort as none is refused. Sending a thinking object with type set to disabled is refused. There is no third way, and no undocumented one that worked.
Know this before you build around it. The usual reason for reaching for those parameters is latency or cost on a simple call, and neither is available to you here. A one-line classification request still thinks first.
The compensating finding is that the effort levels are real. Where many models accept an effort parameter and ignore it, GLM produces measurably different amounts of reasoning at each level. The lever you have is how much it thinks, not whether it thinks, and that lever works.
What to do instead
- If you need a model that answers without reasoning first, use one of the DeepSeek rows instead.
- If you are paying for reasoning you do not want, lower the effort rather than trying to remove it.
How this was measured
An unattributed measurement is a marketing claim, so here is the receipt.
- Probe
- probe_reasoning.py
- Measured
- 2026-09-08
- Applies to
- glm-5.3, glm-5.3-flash
We found this by running the probe against the models we serve, which is also why it is worth trusting: it is a reading rather than a restatement of somebody's documentation. The full catalog is on /models, and what a request costs is on /pricing.
Upstream behaviour changes without notice, and a measurement is only as good as its date. If you find this is no longer true, tell us on the support page and we will re-run the probe.