Context window errors and /compact will not run - Claude Code / Codex
Quick fix
Set the correct model_context_window for the model by hand in cc-switch, so a wrong default does not break compaction or cause a 422.
Symptom
Section titled “Symptom”实际上下文窗口在117.9k就出现了报错且无法/compactThe actual context window errors at 117.9k and /compact will not run
When switching model or generating a model catalogue, cc-switch falls back to a wrong default context window size (1M or 128000) if none is specified explicitly. Claude Code or Codex then cannot calculate the remaining context correctly, so automatic compaction (/compact) fails, it reports full early, or it sends an oversized context upstream and triggers a 422.
The related issues are all still open, so there is no automatic context window sync fix yet — you have to override the default in the config by hand.
In the cc-switch model config, set the correct model_context_window explicitly for the model you use (272000 for GPT-5.6, for example) to override the wrong fallback default.
// cc-switch-model-catalog.json{"models": [{"id": "gpt-5.6-sol","model_context_window": 272000}]}
Affected Versions
Section titled “Affected Versions”Source Issues
Section titled “Source Issues”This page is distilled from 5 real issues
- Why does the context still show 1M after switching model?
- A known cc-switch problem: the context window config is not updated automatically after a model switch, so it still calculates against 1M. Set the correct size by hand.
- Do all models support automatic compaction?
- Not all of them. But if the context window is configured correctly and the model itself supports it, compaction should work — a wrong config breaks it outright.
Related problems
Section titled “Related problems”This is an unofficial community wiki with no affiliation to the cc-switch authors or the project itself. Its content is compiled from the project's public GitHub issues. This site distributes no software.