The Pulse

Anthropic Launches Claude Sonnet 5.5, Claiming 30% Faster Output

Anthropic released Claude Sonnet 5.5 on September 28, 2026, positioning it as a faster, lower-cost model for everyday work and coding. The model generates output more than 30% faster than Sonnet 5 and can cost up to 30% less per task, the c

AI.info Team · Written with AI · fact-checked · 2026-09-28

“Without changing any of our prompts, Claude Sonnet 5.5 did better than Sonnet 5 on almost all of our offline Slackbot evals, in fewer steps and with about 14% fewer output tokens. When someone gives Slackbot a task, quality and speed are what matter most, and Sonnet 5.5 allows Slackbot to deliver better outcomes for users, faster.”

Curtis Allen, Principal Engineer, Slack

Sonnet 5.5 targets everyday work

Anthropic released Claude Sonnet 5.5 on September 28, describing it as the second model in its Claude 5.5 family and a faster, cheaper complement to Claude Opus 5.5. The company is pitching Sonnet for well-scoped tasks such as fixing bugs and preparing documents, slides and spreadsheets, while reserving Opus for work requiring more sustained judgment. Sonnet 5.5 generates output more than 30% faster than Sonnet 5, Anthropic says.

The release also keeps the model’s listed token prices unchanged from Sonnet 5: $2 per million input tokens and $10 per million output tokens. The new model typically uses fewer tokens to complete a task, bringing its cost per task down by as much as 30%, Anthropic says. That distinction matters: the token rates are the same, while the claimed savings come from completing work more efficiently.

A large jump on Anthropic’s coding test

Sonnet 5.5 earns a 70.6% score on Terminal-Bench 4.0, an evaluation of complex, multi-step work in a command-line interface, Anthropic reports. Sonnet 5 scored 10.3% on the same test, according to the company’s comparison. On CursorBench 4.0, which draws on coding tasks from real Cursor sessions, Sonnet 5.5 scored 55.5%, close to Opus 5.5’s 57.8%.

The company also reports gains in knowledge work and computer use. On GDPval-AA, a test spanning 44 occupations and nine industries, Sonnet 5.5 scored 1,844 against Sonnet 5’s 1,449; Opus 5.5 scored 1,846. Those results come from Anthropic’s published comparisons, and the company cautions that benchmark scores capture only part of a model’s capabilities. It also says an external evaluation of some knowledge-work results used a pre-release deployment affected by a structured-output bug, which Anthropic says it has since fixed.

Cybersecurity safeguards accompany the release

Sonnet 5.5’s cybersecurity capabilities are comparable to Opus 5’s, Anthropic says, so it is adding safeguards and fallbacks used for its more capable models. Higher-risk cybersecurity requests can fall back to Sonnet 5, while routine software development remains available. The model also adds safeguards intended to prevent large-scale extraction of its reasoning through distillation attacks.

Sonnet 5.5 is available through Anthropic’s platform, Amazon Web Services, Google Cloud and Microsoft Azure. Developers can call it using the model ID claude-sonnet-5-5. API users who run Sonnet with thinking turned off must switch to the between_tools setting before migrating; the model is also available with zero data retention.

Haiku 5.5 is still to come

Claude Haiku 5.5 will join the family in the coming weeks, Anthropic says, aimed at high-volume and cost-sensitive applications. For now, Sonnet 5.5 sits between that planned model and Opus 5.5: the company’s faster option for coding and routine knowledge work, with the same per-token prices as Sonnet 5 and new controls for higher-risk cybersecurity tasks.