/context shows how much of the context window is in use. The context window is the amount of text the model can take in at once, measured in tokens (small chunks of text). When usage figures from the API are available, the total it shows is now calculated differently. The API is the service Claude Code sends requests to.
Now: the total is the sum of the non-deferred categories plus the remaining message tokens measured by the API.
Before: the total was the raw input token count the API reported.
Why
The headline number is now built from the same categories the breakdown lists, so the total and its parts line up. The number may differ from what earlier versions showed for a similar conversation.
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 how 'non-deferred' categories are chosen or how large the difference from the old total tends to be.
Anthropic's release notes agreeFixed /context total leaving out messages added since the last response; it now matches its categories and can read higher than the status…
