Under the hoodTier: how much it should matter to you
3Useful: my rating, 1 to 5
4Signal: worth watching, 1 to 5
ModelsArea: what it touches
ImprovementsKind: in v2.1.323,
What probably matters to youSection of the release
What
Claude Code decides some behaviour based on what the current model supports, such as whether it accepts a per-turn effort level or tools added mid-conversation. Those decisions used to rely on checks built into Claude Code that look at the model's name. Claude Code now reads a set of capability details for each model from data the server sends, and those override the built-in checks. The /compact context window display uses the same details.
Why
What a model is treated as supporting can be changed from the server without an update to Claude Code.
One source agreesOne thing we can check says the same as this entry.
Anthropic's release notes agreeFixed running /compact again, or after a command such as /context, summarizing away a message that Claude had not answered yet
