Anthropic's New Claude AI Is Faster and Costs Less
Anthropic released Claude Sonnet 5.5 on September 28, 2026, a new version of its middle-sized AI model. The company says it works 30 percent faster than the last version and costs less to use. TechCrunch
Anthropic says the model is built for everyday work, such as writing computer code and drafting office documents. The company says Sonnet 5.5 costs up to 30% less than Sonnet 5 for most work. That is on top of Sonnet 5 prices of $2 for every million pieces of text the model reads and $10 for every million pieces it writes, prices Anthropic has now made permanent. TechCrunch
Anthropic also shares test results for multi-step coding, where the AI does a coding job in stages and uses software tools along the way. Think of it like an assistant who can follow a checklist, use the right programs, and check its own work. On Anthropic's own tests, Sonnet 5.5 scored higher than Opus 5.5, which is the larger and more expensive model.
Anthropic says Sonnet 5.5 has computer-related abilities similar to Opus 5. Because of that, Sonnet 5.5 is the first Sonnet model to follow the same cyber safety rules that already cover Fable and Opus. The company has not described new controls in the materials cited here, only the wider scope.
Sonnet 5.5 arrived less than a week after Opus 5.5. Anthropic released Claude Opus 5.5 on September 22 as the first model in the Claude 5.5 family, with written output priced at $20 per million pieces of text. TechCrunch At that launch, Anthropic said Sonnet 5.5 and Haiku 5.5 would arrive in the coming weeks. BNN Bloomberg Sonnet 5.5 is now out. Anthropic plans to release a new version of Haiku, its smallest model, in the coming weeks.
The broader context here is saving time and money, not only making the AI smarter. When the AI works on its own, it takes many small steps to finish a job. Small savings on each step add up across the whole job.
In my view, technology leaders should test Anthropic's test results on their own projects. Company tests need outside checking. Still, the pattern fits common practice. Many teams save the biggest model for hard problems and use a smaller, faster model for routine work.
Looking at what this means for everyday use, the safety decision is as important as the price. The stricter rules now cover Sonnet, which should simplify company policy but makes careful testing more important. With Opus 5.5 and Sonnet 5.5 out and Haiku 5.5 on the way, users gain a simple option to stay in one model family while adjusting spending per task, which allows more code checked and more drafts written for less time and money.
