Comment by Alifatisk

I used GLM-5.1 when I had the coding plan. Its performance would degrade over time after about 200k tokens. I was suspicious of its recall capability not being that good for long horizon tasks that stress tests the context window. But as they expressed in the tweet:

> It not only supports a truly usable 1M context window but also maintains a continuous lead in the independent completion of long-horizon tasks, providing solid foundational support for building complex agent applications.

Sounds like they have addressed this issue.