Anthropic introduced Claude Sonnet 5.5, the second model in its 5.5 family after the flagship Opus 5.5 shipped a week earlier — a mid-tier release that the company says runs more than 30 percent faster than its predecessor, Sonnet 5, and cuts cost per task by up to 30 percent.
The headline prices do not move. At $2 per million input tokens, $10 per million output tokens and $0.20 per million cache-read tokens, Sonnet 5.5 matches Sonnet 5 exactly; the savings come from needing fewer tokens to finish the same work. That distinction is the whole promise: with list prices flat across the market, token consumption is what shows up on the invoice.
The coding numbers are the eye-catching part. On Terminal-Bench 4.0, Anthropic reports Sonnet 5.5 at 70.6 percent, against 10.3 percent for Sonnet 5 and 66.4 percent for its own flagship Opus 5.5 at the Xhigh setting. On FrontierCode at high compute it scores ten points above Sonnet 5 at the same setting while costing roughly one-fifteenth as much per task. Run at maximum compute, Anthropic says, Sonnet 5.5 matches Opus 5.5 on several tests — but then at similar cost, which removes much of the advantage. The caveat is standard: these are vendor-selected benchmarks, comparing Sonnet 5.5 with Sonnet 5, Opus 5.5 and OpenAI's GPT-6 Sol.
Anthropic positions the model for everyday, well-scoped tasks, bug fixing, and producing documents, slide decks and spreadsheets. It highlights faster iteration on less complex work, more natural communication and improved design sense — the ability to tune a user interface or fill in a presentation template with little rework.
The safety note is notable. Sonnet 5.5 is, per Anthropic, the first Sonnet model to carry guardrails previously reserved for its most capable systems. Because its cybersecurity capabilities are said to be comparable to Opus 5's, requests in that domain are automatically intercepted and routed to a fallback model, while routine software development is left alone. Anthropic concedes the filters will occasionally block legitimate requests too.
The launch lands in the middle of a price and efficiency offensive across the market. OpenAI recently halved prices for GPT-6 Sol and Luna, and per token Sonnet 5.5 now costs the same as GPT-6 Sol. The model is available today on the Claude platform and through Amazon Bedrock, Google Vertex AI and Microsoft Foundry, sitting between Opus 5.5 and the Haiku 5.5 release Anthropic has flagged for the coming weeks.
