Playbooks · Coding tasks
Claude Code 2.1.322: Vertex fast mode, three regions
Written November 13, 2026
A one-change release: Vertex AI fast mode in global, US and EU regions, with a fallback to standard speed when Vertex refuses it.
One change
Claude Code 2.1.322, dated November 10, 2026 in Anthropic's changelog, carries a single entry: fast mode is now available on Vertex AI in the global, us and eu regions, falling back to standard speed if Vertex refuses it. There is nothing else in the release — no second feature hiding under it, and no fix list to work through. Teams on Vertex can take the update for this one behaviour and expect nothing else to move.
Why the region list matters
Fast mode asks the model provider for faster output. Until this release, a team running Claude Code through Google Vertex AI could not count on that option in their deployment region. The changelog now names three: the Vertex global region, plus us and eu. A team pinned to one of those regions for data-residency or procurement reasons gets fast mode inside the deployment they already run, rather than by moving traffic to a different endpoint.
The entry is deliberately narrow. It does not promise fast mode on every Vertex region, and it does not change what fast mode does. It extends an existing option to a provider and a set of regions that did not have it.
The fallback is the operational detail
The second half of the sentence does the quiet work: if Vertex refuses fast mode, the session falls back to standard speed. A refusal does not end the session or surface as a hard failure the user has to clear. Work continues at the normal rate.
That shape matters for how a team rolls this out. Turning fast mode on for a Vertex deployment is not a bet that every request will run faster. Some requests will be refused and will run at standard speed instead, and the tool absorbs that difference without asking anyone to intervene. Cost and latency planning should treat fast mode on Vertex as an improvement when it is granted, not as a guarantee attached to every turn.
What to do
- Update to 2.1.322 or later if your Claude Code sessions run through Vertex AI and you want the option.
- Confirm your Vertex deployment sits in the global, us or eu region; the changelog names no others.
- Expect mixed behaviour in practice: sessions where Vertex grants fast mode run faster, and sessions where it refuses continue at standard speed.
- Nothing else changes in this release, so there is no migration, setting rename, or config edit bundled with it.
How this fits the wider fast-mode picture
Fast mode itself is not new in this release, and 2.1.322 does not redefine it. What changes is where a Vertex-based team can use it. That distinction is worth keeping when you read coverage of this version: reports that describe 2.1.322 as launching fast mode overstate a one-line changelog entry, and reports that dismiss it miss the point for Vertex deployments, where the option genuinely was not on the table in these regions before. For a team already standardized on Vertex, a small release that removes a standing exception — one provider where a familiar option did not apply — is often the kind that matters most in daily use, because it ends the workaround conversations rather than starting new ones.
Limits
- The changelog gives no speed figure for fast mode on Vertex and no list of which models qualify; treat any specific multiplier you see elsewhere as unconfirmed by this release.
- Regions beyond global, us and eu are not named, so a deployment elsewhere should not assume the option exists there.
Related playbooks
- Move fast-mode work off Opus 4.8 in time
- Claude Code 2.1.323: Code Mode writes the tool calls
- Claude Code 2.1.328 makes auto mode the cloud default
- Claude designs three electronic prototypes
Sources
Back to: Coding tasks playbooks hub · Also: Claude Code 2.1.321: Fable 5.2 is default · Claude Code sent your email to websites · Claude Code 2.1.323: Code Mode writes the tool calls