Anthropic launches Claude Sonnet 5.5, 30% faster and up to 30% cheaper than Sonnet 5
2026-09-28
Anthropic released Claude Sonnet 5.5 (claude-sonnet-5-5) on the Claude API, Bedrock, Claude Platform on AWS, Google Cloud, and Microsoft Foundry. Anthropic describes it as 'a clear upgrade over Sonnet 5 that runs 30% faster and costs up to 30% less for most work.' The API changelog documents five breaking behavior changes versus Sonnet 5: a different syntax to disable up-front thinking, a 400 error on forced tool use (tool_choice any/tool), account-scoped thinking blocks, rejection of the older computer_20251124 computer-use tool on the API and Google Cloud, and the advisor tool refusing Opus 4.8/4.7 and Sonnet 5 as advisors.
Significance 3: A mid-tier model refresh with a real but sub-30%-confirmed price/speed improvement and real developer-facing breaking changes, below the frontier-moving bar for a 4.
Implications · machine-drafted, not owner judgment
The breaking changes (tool_choice errors, account-scoped thinking blocks, dropped legacy computer-use tool) mean integrators must actively migrate rather than swap the model string, which slows adoption relative to the price/speed pitch. Anthropic is optimizing its mid-tier line for cost-per-task competitiveness right as OpenAI's GPT-6.1 Sol makes the same pitch on the same day window, intensifying price competition at the mid-tier.
- Independent speed/cost benchmarks (Artificial Analysis) confirming the 30%/30% claims
- Reports of migration friction from the five breaking changes
Sources
- primary Introducing Claude Sonnet 5.5 retrieved 2026-09-30
- primary Claude Platform release notes retrieved 2026-09-30
Affected axes
Output speed · Cost