Claude News
FR
Live

Claude, the essentials — edition of September 30, 2026

Claude Sonnet 5.5 arrives as the default Sonnet, with breaking changes for API users

By ·

Anthropic launched Claude Sonnet 5.5 across its API and the major cloud platforms, and Claude Code 2.1.284 adopted it as the default Sonnet. The release brings API changes that can break code written for Sonnet 5. A separate service incident and three Claude Code releases completed the day.

Key points

Sonnet 5.5 becomes the default Sonnet

Anthropic announced Claude Sonnet 5.5 and published its system card the same day. According to the API release notes dated September 28, 2026, the model, identified as claude-sonnet-5-5, 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 developer guide, "Building with the Claude 5.5 Family", on choosing the right model and getting more from every token.

The figures come from the Claude Code 2.1.284 release notes. They describe Sonnet 5.5 as now the default Sonnet model on the Anthropic API, with a 1M-token context window, a price of $2/$10 per million tokens, and cache reads at $0.20 per million tokens. The same release adds the model to Claude Code. The launch material we reviewed does not detail the model's capabilities, and the system card is the reference for them.

Sources: Introducing Claude Sonnet 5.5 — Anthropic · Claude Sonnet 5.5 System Card — Anthropic · We've launched Claude Sonnet 5.5 (claude-sonnet-5-5) — Anthropic (release notes) · Claude Code 2.1.284 — Claude Code

What developers need to check before migrating

Anthropic's release notes warn that code written for Claude Sonnet 5 can break on Sonnet 5.5 in five ways. Among them, to turn off up-front thinking, developers must send the thinking type between_tools instead of disabled, at high effort or below. Forced tool use, meaning the tool_choice types any and tool, now returns a 400 error. Teams that pin these settings should test before switching the default.

A separate entry in the same notes covers thinking blocks. Those produced by Sonnet 5.5 work only in the account that produced them, or in an account linked to it. When another account sends one, the API drops the block before the model sees it, and the request succeeds. The failure mode is therefore silent: nothing errors, but the block is not used. Applications that pass thinking blocks between accounts are the ones affected.

Sources: Code written for Claude Sonnet 5 can break on Claude Sonnet 5.5 in five ways — Anthropic (release notes) · Thinking blocks that Claude Sonnet 5.5 produces work only in the account that produced them — Anthropic (release notes) · Building with the Claude 5.5 Family: Choosing the Right Model and Getting More from Every Token — Anthropic

Claude Code: three releases centred on control and visibility

Claude Code shipped versions 2.1.283, 2.1.284 and 2.1.285. Version 2.1.283 adds an x-claude-code-prompt-id header to the gateway hint headers, so LLM gateways can group the requests that serve a single user prompt. It is opt-in through CLAUDE_CODE_GATEWAY_HINT_HEADERS=1. Version 2.1.284 shows dollar amounts for the Claude apps gateway spend limit in /usage and in the status line, for example "$271.40 / $500.00 spent this month", when the gateway runs this version or later. The status line's rate_limits.spend_limit also gains used_usd, limit_usd and period.

Version 2.1.284 also 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 such read and still be asked about later ones. Version 2.1.285 adds the CLAUDE_CODE_DISABLE_WEB_FETCH environment variable to turn off the WebFetch tool, the claude --desktop command to open the Claude desktop app on the current directory or on a session with --continue or --resume, and claude plugin configure. That command shows a plugin's options and which are unset, and can save new values read from stdin with --values-stdin.

Sources: Claude Code 2.1.283 — Claude Code · Claude Code 2.1.284 — Claude Code · Claude Code 2.1.285 — Claude Code

A service incident, and two further Anthropic publications

The status page lists an incident titled "Elevated errors on claude.ai, Claude Code, Claude Cowork and the Claude API", now marked resolved. The errors that began at 14:00 UTC were mitigated at 14:36 UTC, and existing conversations were mostly working again. The update also describes a second issue that prevented many users from signing in, since single sign-on and Sign in with Apple were unavailable. It also affected starting new chats, voice conversations, Claude Code or Cowork sessions, purchases and file uploads. Users who were signed in were asked not to sign out. The excerpt we reviewed does not say when the second issue ended.

Anthropic also published two pieces titled "GLM-5.3 and the spread of advanced cyber capabilities" and "What Do You Want from AI?". The material available for this edition carries only their titles, so we do not summarise their content here.

Sources: Claude — Elevated errors on claude.ai, Claude Code, Claude Cowork and the Claude API — status.claude.com · GLM-5.3 and the spread of advanced cyber capabilities — Anthropic · What Do You Want from AI? — Anthropic

This edition is an original synthesis written by Claude from aggregated news — Anthropic's own sources first (release notes, status, newsroom, research, engineering), then the press, Hacker News, Reddit and GitHub, under the editorial supervision of Héra SASU. Every fact links to its article, publisher named. See the live feed →

Claude News is published by Héra SASU. Independent media, not affiliated with Anthropic.