Claude Sonnet 5.5 is now generally available: faster, same price, and stronger security
Anthropic's Claude Sonnet 5.5 runs 30%+ faster at unchanged Sonnet 5 pricing, adds cyber safeguards, and ships with zero data retention across AWS, Google Cloud, and Azure.
Anthropic released Claude Sonnet 5.5 on 28 September 2026, less than a week after Claude Opus 5.5 arrived. It is the second model in the Claude 5.5 family and a direct upgrade to Claude Sonnet 5: faster, more capable for agentic work, and priced identically to what you are already paying.
What has actually changed
The headline figure is speed. Sonnet 5.5 runs more than 30% faster than Sonnet 5, and the efficiency gain comes from a meaningful architectural change rather than a brute-force speed-up. The model completes tasks using fewer tokens and fewer tool calls. Lovable, one of the early testers, reported roughly a third fewer tool calls in coding tests. That matters because in agentic workflows, tool calls accumulate cost quickly. Getting the same result with fewer of them is where the real saving lands.
The pricing stays at $2 per million input tokens and $10 per million output tokens, with cache reads at $0.20 per million. If you are already on Sonnet 5, switching to Sonnet 5.5 costs you nothing extra and should cost you noticeably less across longer tasks.
Benchmark scores give some sense of the capability jump. On Terminal-Bench 4.0, an agentic coding evaluation, Sonnet 5.5 scores 70.6%. Sonnet 5 scored 10.3% on the same benchmark. Opus 5.5, the more powerful model in the family, scores 66.4%, meaning Sonnet 5.5 actually comes out ahead there. On CursorBench 4.0 it scores 55.5%, compared with 34.1% for Sonnet 5 and 57.8% for Opus 5.5. These are not small incremental improvements.
Anthropic also notes that Sonnet 5.5 is the first Sonnet model to complete Pokémon Red working only from screenshots, which is a proxy for strong long-horizon reasoning and image understanding rather than a party trick.
The security upgrade that matters for enterprises
Sonnet 5.5 is the first model in the Sonnet line to ship with cybersecurity safeguards comparable to those Anthropic previously reserved for Opus-class models. This is new territory for the mid-tier family.
Because the model is significantly more capable in cybersecurity than its predecessor, Anthropic has added a visible fallback mechanism: requests that involve high-risk cybersecurity tasks are rerouted to Sonnet 5. Qualified professionals can apply for tiered access through Anthropic’s expanded Cyber Verification Programme.
On top of that, Sonnet 5.5 is the first Sonnet model to use safety classifiers specifically designed to prevent distillation attacks, where someone tries to extract a model’s reasoning at scale. Invisible text watermarking is also included. In an automated audit covering approximately 1,850 scenarios, the model matched or improved on Sonnet 5 across measures of alignment, resistance to misuse, and honesty. Anthropic found no evidence of goal-seeking behaviour that conflicted with user intentions.
Zero data retention is available across Amazon Web Services, Google Cloud, and Microsoft Azure. Anthropic has committed not to retire the model before 27 September 2027.
What this means for you
If you are a developer using the API, the model ID is claude-sonnet-5-5. One migration note worth flagging: if you run Sonnet with thinking turned off, you will need to switch to the new between_tools setting before moving to Sonnet 5.5. The migration guide covers this. In Claude Code, the sonnet alias switches to 5.5 from version 2.1.284 onwards, but Opus 5.5 remains the default so you need to select Sonnet explicitly.
Anthropic has also recalibrated five effort levels: low, medium, high, xhigh, and max. Lower settings favour speed and cost; higher ones allow longer reasoning. The default is medium in the Claude apps and Claude Code, and high on the API.
If you use GitHub Copilot, Sonnet 5.5 is now generally available for Pro, Pro+, Max, Business, and Enterprise users. It is designed for well-scoped everyday work such as building features and fixing bugs.
If you are evaluating this for enterprise use, Daniel Vogel, chief operating officer at Epic Games, is among the early testers cited by Anthropic. He reported that the model handled tens of thousands of lines of gameplay-system code alongside system design and data-flow reviews without issue.
If your organisation has compliance requirements, zero data retention is available on all three major cloud platforms from day one. On AWS Bedrock specifically, the model integrates with AWS IAM, AWS CloudTrail, Amazon CloudWatch, and Amazon Bedrock Guardrails, with data staying within AWS infrastructure and regional data residency options intact.
What is coming next
Claude Haiku 5.5 is the third model planned for the Claude 5.5 family. It is built for high-volume and cost-sensitive applications and is expected in the weeks following this release. No specific date has been confirmed.
Sonnet 5.5 sits in a clear position in the Claude family. Opus 5.5 handles complex work requiring careful judgement. Sonnet 5.5 covers well-scoped everyday tasks, bug fixing, and producing polished documents, slides, and spreadsheets. Haiku 5.5 will cover the high-volume, lower-cost end.
For most teams currently running Sonnet 5 workloads, the upgrade path is straightforward and the economics are better. The security additions mean it is also now viable for organisations that previously had to use Opus-class models purely for compliance or safety reasons.