Follow Discord
Sweep 25 Sep 2026 · 19:33Z Build v2.1.283 504 read Stable v2.1.274 Latest v2.1.283 Next v2.1.283 Feeds RSS JSON llms.txt llms-full.txt Unofficial

Claude Code v2.1.281 ·

Context-low warning passes the model through a normalizer

The low-context warning now normalizes the model before working out its warning level and auto-compact threshold

Entry
JSON All of v2.1.281
EntryKind: in v2.1.281,
ChangesSection of the release
What

The "Context low" warning and the "% until auto-compact" footer now run the current model through a helper first. From the result they work out three things: the warning level, the usable context size and the point where auto-compact starts. Context is how much of the conversation the model can hold at once. Auto-compact is when Claude Code shortens the conversation to make room. The wording of the warning is unchanged.

Why

The warning's numbers now come from the model after this extra step, so the percentage shown can differ for some models even though the text looks the same.

How sure we are
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubtThe finding does not say what the helper does to the model, so it is unclear for which models the numbers change.

See this entry in the whole of v2.1.281 →

Feedback