Follow Discord
Sweep 29 Sep 2026 · 18:10Z Build v2.1.285 506 read Stable v2.1.280 Latest v2.1.285 Next v2.1.285 Feeds RSS JSON llms.txt llms-full.txt Unofficial

Claude Code v2.1.285 ·

Bedrock and Vertex AI model checks remember results and recheck before falling back

On Bedrock and Vertex AI, Claude Code now remembers per-model access results with expiry and rechecks a model live before using it

You'll notice Improvements
JSON All of v2.1.285
You'll noticeTier: how much it should matter to you
2Useful: my rating, 1 to 5
1Signal: worth watching, 1 to 5
ModelsArea: what it touches
ImprovementsKind: in v2.1.285,
ImprovementsSection of the release
What

When you use Claude through Amazon Bedrock or Google Vertex AI and have not pinned a specific model, Claude Code checks which models your account can use and falls back to another if needed. Before, it kept a single cached yes-or-no answer. Now:

  • Each model gets its own recorded result: refused, denied, temporary or accessible, and each result expires.
  • Models that refused are remembered.
  • A candidate model is confirmed with a fresh live check before it is used.
  • A version check can mark a model as unavailable.
  • Saved results are tied to the endpoint address and to whether authentication is skipped, so changing either does not reuse old results.

None of this runs when CLAUDE_CODE_PROVIDER_MANAGED_BY_HOST is set.

Why

If you use Bedrock or Vertex AI, you should see fewer false "model unavailable" fallbacks and fewer repeated checks.

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 doubtHow long each remembered result lasts, and exactly what you will see differently, is not stated.

See this entry in the whole of v2.1.285 →

Feedback