Claude Sonnet 5.5 arrives as the default Sonnet, with migration caveats for developers

Sonnet 5.5 launches with a system card and a new default

Anthropic introduced Claude Sonnet 5.5 and published a system card alongside the announcement. According to Claude Code's 2.1.284 release notes, the model (claude-sonnet-5-5) is now the default Sonnet model on the Anthropic API. It offers a 1M-token context window and is priced at $2 per million input tokens and $10 per million output tokens, with cache reads at $0.20 per million tokens.

The API release notes, dated September 28, say the model is available on the Claude API, Claude in Amazon Bedrock, Claude Platform on AWS, Claude on Google Cloud and Claude in Microsoft Foundry. Anthropic also published a page titled "Opus 5.5 for Work" on the same day. The provided item gives no detail beyond its title.

Sources: Introducing Claude Sonnet 5.5 — Anthropic · Claude Sonnet 5.5 System Card — Anthropic · Claude API release notes, September 28, 2026 — Anthropic (release notes) · Claude Code 2.1.284 — Claude Code

What developers need to check before switching

Anthropic's release notes warn that code written for Claude Sonnet 5 can break on Sonnet 5.5 in five ways. Two are spelled out in the provided excerpts. To turn off up-front thinking, requests must send thinking: {"type": "between_tools"} instead of "disabled", at high effort or below. Forced tool use, meaning tool_choice types any and tool, now returns a 400 error. The excerpt does not list the other three.

A further change concerns thinking blocks. Those produced by Sonnet 5.5 work only in the account that produced them, or in an account linked to it. If another account sends such a block, the API drops it before the model sees it, and the request still succeeds. Because the request succeeds silently, teams that pass thinking blocks between accounts should test for this. The notes say blocks from earlier models are treated differently, but the excerpt is cut off before the detail.

Sources: Claude API release notes, September 28, 2026 — Anthropic (release notes)

Claude Code adds model controls and cost visibility

Claude Code 2.1.283 gives administrators more control over which models are available. A new availableModelsMatch managed setting, set to "exact", makes an availableModels entry allow only the model version it names, so new releases stay blocked until they are listed. A separate deniedModels setting blocks specific models even when availableModels allows them. The release also adds an x-claude-code-prompt-id gateway hint header, enabled with CLAUDE_CODE_GATEWAY_HINT_HEADERS=1, so LLM gateways can group the requests that serve one user prompt.

Version 2.1.284 adds a "Yes, but ask again next time" answer to auto mode's prompt before a read outside the working directories. It lets a user allow one read and still be asked about later ones. It also shows dollar amounts for the Claude apps gateway spend limit in /usage and the status line, for example "$271.40 / $500.00 spent this month", when the gateway runs this version or later. Version 2.1.282 added a maxProseWidth setting for wide terminals, plus a startup notice and /status and claude doctor entries listing telemetry variables in project settings files that were ignored or that turned telemetry off.

Sources: Claude Code 2.1.283 — Claude Code · Claude Code 2.1.284 — Claude Code · Claude Code 2.1.282 — Claude Code

Research: a nine-loop amplitude computed with Claude

Anthropic also published a research post titled "Yes, Claude can do Nine Loops", accompanied by an announcement that Claude computes a nine-loop amplitude in N=4 super-Yang-Mills. The provided items contain only these titles, with no summary. Details of the method, the result and any human involvement therefore cannot be reported here.

Sources: Yes, Claude can do Nine Loops — Anthropic (research) · Claude computes a nine-loop amplitude in N=4 super-Yang-Mills — Anthropic