Anthropic’s Claude Sonnet 5.5 is aimed squarely at the everyday work companies actually need AI to perform: coding, document creation, analysis and repeatable knowledge work — while improving speed and lowering the cost per completed task.
Anthropic says Sonnet 5.5 runs more than 30% faster than Sonnet 5 and can cost up to 30% less per task, despite keeping the same headline token prices.
Claude Sonnet 5.5 arrives
Anthropic announced Claude Sonnet 5.5 on 28 September 2026 as the second model in its Claude 5.5 family. The company positions it below the more expensive Opus 5.5 for the hardest open-ended work, but says Sonnet 5.5 is particularly strong at well-scoped everyday tasks, bug fixing, coding and the creation of polished documents, presentations and spreadsheets.
The emphasis on completed work rather than only chat quality is notable. Enterprise AI increasingly lives inside processes where the model has to understand source material, use tools, manipulate files and return an output that is ready for a person to use.
A large jump in coding performance
Anthropic reports substantial gains over Sonnet 5 across several coding evaluations. It says Sonnet 5.5 reached 70.6% on Terminal-Bench 4.0, compared with 10.3% for Sonnet 5, and performed close to Opus 5.5 on a number of professional-work benchmarks.
The company also reports stronger long-horizon work, computer use, chart recognition and visual understanding. Early enterprise testers cited faster ticket handling, fewer tool calls and lower token consumption in their own internal evaluations.
Those figures are useful signals, but businesses should still run controlled tests on their own work. A customer-support workflow, a software repository and a financial-document process can place very different demands on the same model.
Why “same token price” can still mean lower cost
Anthropic kept Sonnet 5.5’s list price at US$2 per million input tokens and US$10 per million output tokens. The company says the model generally needs fewer tokens and fewer steps to finish the same work, which can reduce the cost per task by up to 30% in its testing.
This distinction matters when budgeting for AI. The number that matters most is often not the price of a million tokens; it is the cost of reliably completing one business outcome. Faster execution, fewer retries and less human correction can materially change that calculation.
Where businesses may feel the difference
Sonnet 5.5’s profile makes it relevant to high-volume professional workflows: analysing documents, preparing presentations, assisting developers, handling support tickets, summarising operational information and working through repetitive knowledge tasks.
Anthropic also highlights design and document quality. That could make models in this class increasingly useful not only for text generation but for producing client-facing deliverables that need less manual cleanup.
The latest Claude release reinforces a broader trend: the useful AI stack is becoming faster and cheaper while models get better at tools and multi-step execution. For SMEs, this opens the door to more practical automation — but only when permissions, human escalation, data privacy and process design are handled properly.
Availability and safeguards
Anthropic says Claude Sonnet 5.5 is available across its platforms and through major cloud providers. The model also launches with stronger cybersecurity safeguards than the previous Sonnet generation, reflecting the increase in its technical capabilities.
