Model switches return error codes and skip the server check for…

What

A failed model switch now returns a machine-readable code. The code is one of alias_1m_disabled, alias_1m_unsupported, restricted_by_org, unavailable_for_account, not_offered, check_failed or, when a hook blocked the switch, blocked_by_hook. A hook is a command you configure to run on events.
Error messages mention /model only where that command exists. Elsewhere, they say "Switch to a different model." or just "Try again." Before, they always said to run /model.
The "Couldn't confirm model" error is now marked retryable.
A model picked at launch, for example by the SDK or in print mode (-p), is no longer checked against the server. The switch passes confirmWithServer: false, which becomes serverProbe: false, and the model is accepted as valid. The server check is also skipped when no deadline is given.
The print and SDK session now reports when a pick is a launch pick (isLaunchModelPick).

Why

The SDK is the library for running Claude Code from your own programs, and programs that embed Claude Code often have no /model command. Those hosts now get messages that make sense and a code they can act on. Starting with a model chosen at launch no longer waits on a server check, which matters most on a slow network.

Read from

Names in the bundle/model

What the documentation says

/model Connect Claude Desktop to Claude for Government modified, medium confidence * New and retired models appear in the model picker without any profile change; model access is controlled through [seat tiers](/docs/government/org-admin/seat-tiers). [Models in Claude Desktop](/docs/government/desktop/models) covers a mo… see the edit

Since it was published

The entry above is what we published on the day. These lines were added later, as Anthropic's own pages caught up, and they sit beside the original rather than replacing it.

Confirmed since 24 Sep 2026 Anthropic's documentation has since written up /model, on optimizing-for-cost-and-intelligence. | You want agent runs to finish sooner | Tell the model that time matters, and show it the elapsed time; on DRACO, HLE, and an internal physics set, runs took 33% to 69% less time at a 28% to 54% lower cost per task, with scores up to 1.9 … about-claude/models/optimizing-for-cost-and-intelligence see the edit

Added since 24 Sep 2026 A small documentation edit on Deploy managed settings touched a line naming /model after this was published. Wherever the session runs, claude.ai applies the admin console's [`strictKnownMarketplaces`](/docs/en/settings-reference#strictknownmarketplaces) and [`blockedMarketplaces`](/docs/en/settings-reference#blockedmarketplaces) lists itself whe… managed-settings see the edit

How sure we are

Two sources agreeTwo things we can check say the same as this entry.

Anthropic's documentation agrees/model on Connect Claude Desktop to Claude for Government

Anthropic's release notes agreeFixed /model showing the raw API error JSON and request ID when the API refuses the picked model; it now shows the server's message and says…

See this entry in the whole of v2.1.281 →

The build it came inv2.1.281

23 Sep 2026 · 1,065 entries, this one is #89.

Use it now39
You'll notice261
Nothing to try yet26
Under the hood217

Either side of it

v2.1.280Older
v2.1.282Newer

Open the release →

Names in this entry

slash/model
Switch the AI model and save it as your default for new sessions. For models that support it, use left/right arrows to adjust effort level. With no argument, opens a picker; press s on a row to switch for the current session only. See when Claude Code asks you to confirm the switch. Once you confirm the switch, if Claude Code asks, Claude Code applies the change without waiting for the current response to finish. Before v2.1.242, Claude Code decided from a feature flag it fetched from Anthropic whether to run the command mid-turn or queue it until the turn finished, and always queued it in a session that doesn't fetch feature flags, such as on a third-party provider. Also available in non-interactive mode (-p) with a model argument instead of the picker, where it applies to the current session only and isn't saved as your default; requires Claude Code v2.1.205 or later

Anthropic's note

Upstream's own release notes for v2.1.281 name something this entry names, so the two are probably about the same change. Probably: the pairing is made on a shared name rather than on the words around it.

Fixed /model showing the raw API error JSON and request ID when the API refuses the picked model; it now shows the server's message and says the model was not changed

Is this right?