Anthropic just dropped a brand-new AI model: Claude Sonnet 5.5, released on Monday, September 28, 2026. It's the second model in the Claude 5.5 family — Claude Opus 5.5 launched the week before. Anthropic calls it a "clear upgrade" over Sonnet 5, and the headline is hard to miss: output that's more than 30% faster and costs up to 30% less for most work.
Faster and cheaper — where do the savings come from?
The pricing details tell an interesting story. Per-token pricing is unchanged from Sonnet 5: $2 per million input tokens, $10 per million output tokens, and $0.20 per million cache reads. So where does the "up to 30% cheaper" come from? Efficiency. Sonnet 5.5 needs far fewer tokens to finish the same job, so the savings show up on your bill rather than on the price list.
A coding powerhouse
The benchmark numbers are striking. Sonnet 5.5 scores 70.6% on Terminal-Bench 4.0 — a giant leap from Sonnet 5's 10.3%, and even above the flagship Opus 5.5's 66.4%. It hits 55.5% on CursorBench 4.0 (up from 34.1%) and beats Sonnet 5 by ten points on FrontierCode 1.1 at one-fifteenth of the task cost. It's also the first Sonnet to beat Nintendo's Pokémon Red working only from screenshots. For everyday work, Anthropic says it shines at well-scoped tasks: fixing bugs and writing polished documents, slides, and spreadsheets. It has a sharp eye for design, writes more clearly than Sonnet 5, and is the first Sonnet to ship with cybersecurity safeguards comparable to Anthropic's most capable models.
Where to get it — and what comes next
Sonnet 5.5 is available now on the Claude Platform (model id: claude-sonnet-5-5) and on AWS, Google Cloud, and Microsoft Azure. The 5.5 family isn't complete yet, though — a smaller, high-throughput Haiku 5.5 arrives in the coming weeks.
The bigger picture
Some context helps explain the timing. Anthropic says enterprise customers make up around 80% of its business, with clients like Salesforce, Databricks, Goldman Sachs, and Novo Nordisk — and the company is expanding ahead of a planned IPO. What makes this launch especially notable is the timing: earlier this month, CEO Dario Amodei publicly called on the AI community to slow the pace of releases over safety concerns. A fast follow-up launch just weeks later is hard to ignore. Still, the short take is simple: the mid-tier model now beats the flagship at coding for half the price. The AI race is shifting from "smartest" to "smartest per dollar."
Comments
Post a Comment