Some GLM-4.6 (reasoning) responses cut off very quickly

foxberg34

If I ask a question to GLM-4.6 (reasoning) that involves math, the response produces a short "thinking" block and then stops with this message:

======The response is incomplete because it exceeded the model's output limit.

Example thread:
https://kagi.com/assistant/e9f6caf9-463e-4c8d-a93a-ca9a56062d81

Strangely, asking open-ended non-math questions seems to consistently work well, even if they also produce a long thinking prompt.

RoxyRoxyRoxy

CC @yiwei-1

yiwei-1

This is probably a provider-side issue, as we don’t have special output limits for GLM 4.6 models. We’ll be updating to GLM 4.7 models soon, which should improve overall reliability. Please let us know if you still encounter this issue after the update.

foxberg34

yiwei-1 GLM-4.7 seems to work fine.