Anthropic launches Claude Sonnet 5.5, 30% faster at unchanged token prices

Anthropic released Claude Sonnet 5.5 on 28 September 2026, describing it as the second model in the Claude 5.5 family. The company says the mid-tier model runs more than 30% faster than Sonnet 5 and cuts cost per task by up to 30%, while keeping the predecessor's token prices unchanged. It follows Opus 5.5, which arrived the previous week.thenewstack.io

The release matters because Sonnet sits below Opus in Anthropic's lineup yet now approaches it on coding benchmarks at a lower effective cost. Anthropic's own benchmarks show Sonnet 5.5 outperforming Opus 5.5 on agentic coding, which the company attributes to its ability to spawn multiple agents without exceeding cost limits. Sonnet 5, announced about three months earlier, had sold on efficient agentic deployment at lower cost than competitors.techcrunch.com

What the release changes

Anthropic reports Sonnet 5.5 runs more than 30% faster than Sonnet 5 and typically uses far fewer tokens to complete the same work, making it the fastest Sonnet model to date. Pricing is unchanged: $2 per million input tokens, $10 per million output tokens, and $0.20 per million cache reads. Because fewer tokens are needed, effective cost per task falls by up to 30%.9to5mac.com

On Terminal-Bench 4.0, a coding benchmark, Sonnet 5.5 reaches 70.6%, against 10.3% for Sonnet 5 and 66.4% (Xhigh) for Opus 5.5, according to Anthropic's announcement. On FrontierCode, at high compute it scores 10 points more than Sonnet 5 at the same setting, at roughly one-fifteenth the cost per task. Heise notes Anthropic presents only selected benchmarks, comparing Sonnet 5.5 with its predecessor, the next-larger model, and OpenAI's GPT-6 Sol.

Sonnet 5.5 is positioned between the released flagship Opus 5.5 and the Haiku 5.5 announced for the coming weeks. Per token it costs the same as OpenAI's GPT-6 Sol, also at $2 and $10. The launch comes amid a price and efficiency push in the AI market: OpenAI recently halved prices for GPT-6 Sol and Luna. Anthropic also plans a new Haiku version without a firm date.

Where the benchmark numbers came from

Anthropic says Sonnet 5.5 is strongest at everyday tasks with a clear scope: fixing bugs and creating documents, slides and spreadsheets. The company highlights design work, from refining user interfaces to filling presentation templates with minimal editing. Early testers described it as a better collaborator than Sonnet 5, with clearer writing and speed suited to quick iteration on less complex tasks, Anthropic says.

Sonnet 5.5 has cybersecurity capabilities Anthropic calls comparable to Opus 5, making it the first Sonnet model to launch with cyber safeguards and fallbacks similar to those used for its most capable models. Risky requests are intercepted and routed to a fallback model, while routine software development stays unaffected. Anthropic acknowledges the filters can occasionally block legitimate requests. Biology safeguards remain unchanged from Sonnet 5.

Sonnet 5.5 is available immediately on the Claude platform, Amazon Bedrock, Google Vertex AI and Microsoft Foundry. Anthropic has also put the model live in Claude Code. The company says both its cyber and biology safeguards target a narrow set of high-risk requests, leaving routine software development and most life sciences work unaffected. Anthropic has not given a firm date for the new Haiku version, saying only that it will arrive in the coming weeks.