Source Intelligence
Follow Discord

DisclaimerUnofficial, and not affiliated with Anthropic. Nearly all of this is read straight out of what ships: npm bundles, captured prompts, published docs. Anthropic's own notes go in verbatim, marked as theirs. The rest is my reading, and every entry carries the strings behind it. If one looks wrong, vote it down and say why.

← All of v2.1.234 Home All releases ←olderv2.1.233 v2.1.235newer→
Claude Code v2.1.234

Per-model token limits fetched from the API

Under the hood
Useful2 Signal3
Model Config

Model token limits can now come from the API with caching and a timeout instead of a built-in table.

What

Claude Code can now read a model's input and output token limits from the API rather than a built-in table, caching the result per model id and timing the request out rather than blocking a turn indefinitely.

Details
  • Failures are classified as timeout, not_found, http_<status> or connection.
  • A separate telemetry path records the case where a turn had to wait on the fetch instead of using a cached value.
  • Reached through an internal predicate rather than a named flag, so whether it runs is decided outside this file.
Evidence

dynamic model config fetch timed out

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Related

Other releases about the same thing. Found by shared names or similar wording; neither means one caused the other.

See this entry in the whole of v2.1.234 →