Anthropic launches Claude Sonnet 5.5, a faster and cheaper mid-range model
2026-09-28
Anthropic released Claude Sonnet 5.5 (claude-sonnet-5-5) on the Claude API, Amazon Bedrock, Google Cloud, and Microsoft Foundry. Per Anthropic's own announcement and TechCrunch, it runs about 30% faster and costs up to 30% less than Sonnet 5 for most work. Anthropic's API changelog also documents several breaking changes for code written against Sonnet 5: a changed parameter for disabling up-front thinking, a 400 error on forced tool use (tool_choice types any/tool), rejection of the prior computer_20251124 computer-use tool, and thinking blocks that are now tied to the specific account that generated them.
Significance 4: A new flagship mid-range model with an up-to-30% price cut and speed improvement is a real pricing_change plus release event, capped at 4 because the performance and cost claims are lab-reported until independently verified.
Implications · machine-drafted, not owner judgment
A cheaper, faster mid-tier Claude model narrows the cost gap that has been part of the daily-driver and bulk-cheap calculus, and gives Anthropic a sharper answer to GPT-5.2's earlier price cut. For agent-heavy workloads (coding agents, always-on personal agents like Muse or Instinct) where token volume dominates cost, a 30% price/speed improvement on a mid-tier model compounds quickly and could shift where builders default.
- Independent cost/speed benchmarks (Artificial Analysis) confirming the 30% figures
- Whether Sonnet 5.5 shows up as the default model in agent harnesses (Claude Code, Cursor) within a month
Sources
- primary Claude Platform release notes retrieved 2026-09-28
- primary Introducing Claude Sonnet 5.5 retrieved 2026-09-28
- press Anthropic releases Sonnet 5.5, which it calls a significantly cheaper, faster work partner retrieved 2026-09-28
Affected axes
Cost · Output speed