What probably matters to youSection of the release
Unclear It is not clear what decides whether Sonnet 5 counts as available.
What
A new model, Sonnet 5.5 (claude-sonnet-5-5, display name "Sonnet 5.5", 1M context window), is added and becomes the default for the sonnet family when you use Anthropic's own API. A context window is how much text the model can hold at once.
A short name sonnet-5-5 maps to claude-sonnet-5-5, checked ahead of sonnet-5 so the new ID is no longer treated as Sonnet 5. It is added to the known-models list and gets its own Vertex region environment variable.
Other providers keep their own Sonnet defaults: Bedrock, Vertex, Foundry and Mantle stay on claude-sonnet-4-5, and anthropic_aws and gateway stay on claude-sonnet-4-6.
The /model picker gains a row labelled "Sonnet 5 · Previous Sonnet version" when Sonnet 5 is available.
Internal model names move on: the main Opus and Sonnet IDs are now claude-opus-5-5 and claude-sonnet-5-5 ("Claude Sonnet 5.5"), the previous ones are claude-opus-5 and claude-sonnet-5, and the separate "next" entries are gone.
Refusal fallback: when a Sonnet 5.5 session (claude-sonnet-5-5 or claude-sonnet-5-5[1m]) has a request refused in the cyber or frontier_llm category, Claude Code falls back to claude-sonnet-5. There is no fallback for bio. Before, Sonnet 5.5 would have used the default list, which sends bio and frontier_llm to claude-opus-5 and cyber to claude-opus-4-8.
Safeguard notices: the "You may be seeing this for the first time…" explanation, shown when safeguards flag a message, used to appear only for claude-opus-5-5. It now also appears for any Sonnet model in the cyber category, reading "on a Sonnet model". Sonnet also gets the frontier_llm recovery text ("is answering instead" / "Edit and retry"). The notices now come from a per-family table, and the Opus wording is built the same way instead of being fixed text.
Why
If you pick sonnet on the Anthropic API, you move to a new model without changing anything, and you can still choose Sonnet 5 from the picker. Sonnet users who hit a safeguard refusal now get an explanation and a fallback route instead of a bare refusal notice.
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.
Added sinceA small documentation edit on Claude Code on Amazon Bedrock touched a line naming /model after this was published.The [setup wizard](#sign-in-with-bedrock) offers a 1M context option when it pins models. To enable it for a manually pinned model instead, append `[1m]` to the model ID. See [Pin models for third-party deployments](/docs/en/model-config#p…amazon-bedrocksee the edit
Confirmed sinceAnthropic's documentation has since written up /model, on spawn.sh: called once per claimed work item.The SDK's [Client-side MCP helpers](https://platform.claude.com/docs/en/agents-and-tools/mcp-connector#client-side-mcp-helpers) convert the server's tools into the runnable tools the worker accepts. Install an MCP SDK alongside the Anthrop…managed-agents/self-hosted-sandboxessee the edit
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 doubtIt is not clear what decides whether Sonnet 5 counts as available.
Anthropic's documentation agreesAnthropic's documentation has since written up /model, on spawn.sh: called once per claimed work item.
Anthropic's release notes agreeFixed the /model picker showing "+1 model" below the list after scrolling to the last model; the count now covers only the models below the…