Context auto-compression seems ineffective with GLM-5: Request exceeds maximum context length

[mafamily2496]

Description

I am using opencode with the GLM-5 model. According to the features/documentation, opencode should support automatic context compression/summarization when the conversation history or input size grows large.

However, I encountered an API error indicating that the input token count exceeds the model's limit, suggesting that the auto-compression mechanism either did not trigger or failed to reduce the context sufficiently.

Plugins

No response

OpenCode version

1.1.15

Steps to reproduce

Initialize opencode with the GLM-5 model configuration.
Start a long-running conversation session (e.g., iterating on a large codebase or analyzing multiple files).
Monitor the context usage until it reaches approximately 170,000 tokens.
At this stage, trigger an action that requires tool usage (e.g., file edit, search) or subagent invocation.
Observe the API error: Request ... input tokens exceeds the model's maximum context length.

Screenshot and/or share link

No response

Operating System

windows 11

Terminal

windows terminal

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• Auto-compaction enabled by default, but context_length_exceeded errors still occur in agent workflows #8089: Auto-compaction enabled by default, but context_length_exceeded errors still occur in agent workflows (very similar symptom)
• Compaction overflow check doesn't account for large tool outputs until the next step, causing context overflow #10634: Compaction overflow check doesn't account for large tool outputs until the next step, causing context overflow (related root cause)
• Compressing Chat don t work properly well after so many Updates from OpenCode #12380: Compressing Chat don't work properly well after so many Updates from OpenCode (related)
• [BUG]: Auto compaction not triggering reliably with Anthropic models #6068: [BUG]: Auto compaction not triggering reliably with Anthropic models (same symptom, different model)

If your issue is distinct (e.g., specific to the GLM-5 model or a different failure mode), please add details clarifying how it differs from the above.