AI Change Tracker
3

Anthropic launches Claude Sonnet 5.5, a cheaper/faster Sonnet-tier model with breaking API changes

2026-09-28

On September 28, 2026, Anthropic launched Claude Sonnet 5.5 (claude-sonnet-5-5) on the Claude API, Amazon Bedrock, Claude Platform on AWS, Claude on Google Cloud, and Claude in Microsoft Foundry. Anthropic's announcement states it "runs 30% faster and costs up to 30% less for most work" than Sonnet 5 at the same list price. The API changelog documents five breaking changes for existing Sonnet 5 code (thinking-parameter syntax, forced tool_choice now returning a 400 error, account-scoped thinking blocks, a dropped computer-use tool version, and advisor-tool model rejections) and extends the 'preserved thinking' anti-distillation mechanism. Sonnet 5.5 now powers the free tier on claude.ai; per Simon Willison, ChatGPT's free tier still runs the older GPT-5.6 Luna, which he calls "a lot less capable." Third-party availability includes GitHub Copilot, Cursor, Factory, Devin, and Cline.

Significance 3: An iterative Sonnet-tier release with real but sub-30%-cap pricing/speed gains and several breaking API changes; it doesn't unseat Opus 4.8/Opus 5 on tracked verdicts, but the free-tier swap is a meaningful consumer-facing move against ChatGPT's free tier.

technology financial business Release Pricing change Tooling / agents

Implications · machine-drafted, not owner judgment

Anthropic is shipping on a near-weekly cadence across the 5.5 family (Opus 5.5 then Sonnet 5.5 within a week), pressuring OpenAI and Google to match cadence rather than just capability, and giving Anthropic a materially stronger free-tier offering than ChatGPT's heading into OpenAI's DevDay. The anti-distillation 'preserved thinking' change also signals Anthropic is actively hardening against reasoning-trace extraction, consistent with its own distillation allegations against Chinese labs.

Watch for
  • Independent benchmark placement of Sonnet 5.5 relative to Opus 5.5 and GPT-6 Astra/Sol within a few weeks
  • Any measurable shift in claude.ai vs. ChatGPT free-tier usage

Sources

Affected axes

Cost · Output speed

← Back to feed