SiliconFlow GLM-5.2 context/output limits incorrect

[Sequlumine]

GLM-5.2 has incorrect limit values in both SiliconFlow providers. Using an incorrect number of contexts causes the input to be compressed prematurely during use.

Files:

• providers/siliconflow/models/zai-org/GLM-5.2.toml
• providers/siliconflow-cn/models/zai-org/GLM-5.2.toml

Current:
[limit] context = 205_000 output = 205_000

Expected:
[limit] context = 1_049_000 output = 262_000

https://www.siliconflow.com/models/glm-5-2

[imagebuilder1837]

I’m seeing the same issue in OpenCode with siliconflow-cn/zai-org/GLM-5.2. The 205K provider limit causes premature compaction, even though GLM-5.2 is advertised as a 1M-context model. Updating the SiliconFlow entries would be very helpful.