Five releases in three days — Anthropic’s shipping pace for Claude Code remains remarkable. Between June 4 and 6, versions 2.1.163 through 2.1.167 landed. The headline feature: an intelligent fallback system for models.
The fallback system
Version 2.1.166 introduces a new fallbackModel setting. You can now configure up to three backup models that are tried in order when the primary model is overloaded or unavailable. The --fallback-model option now works in interactive sessions too.
Version 2.1.167 extends this: Claude Code automatically retries a turn once on the fallback model when the API rejects with an unexpected non-retryable error. Auth, rate-limit, and transport errors still surface immediately.
Other highlights
Version 2.1.163 (June 4) was a substantial update with numerous fixes: better startup reliability, fixed keyboard issues with Kitty terminals, more stable background agent sessions, and cleaner menus. Background agent sessions now update to new Claude Code versions in the background, so opening a session after an update no longer waits on a cold restart.
A nice quality-of-life change: clicking a slash command in the autocomplete menu now fills it into your prompt instead of running it immediately — a small detail that prevents a lot of accidental executions.
My take
The fallback system is the right answer to a real problem. If you use Claude Code daily, you know the frustration: the model is overloaded, and you wait. With fallback, Claude Code automatically switches to a backup model — no manual intervention needed. In a world where model routing is becoming a trend, this is a smart move.
Sources: Claude Code Changelog, GitHub Releases