Anthropic said Sonnet 5.5 cut task costs up to 30%.
Anthropic released Claude Sonnet 5.5 on 28 September at the same price as Sonnet 5: $2 per million input tokens and $10 per million output tokens.
Its claim is that the model needs fewer tokens to finish the same job. In Anthropic's testing, that reduced task costs by up to 30% and produced results more than 30% faster.
The largest reported jump came on Terminal-Bench 4.0, which tests autonomous coding work. Sonnet 5.5 scored 70.6%, compared with 10.3% for Sonnet 5.
This is also the first Sonnet release using the cybersecurity safeguards Anthropic applies to its most capable models. The company says its cyber capabilities are comparable to Opus 5, although it says Opus 5.5 remains stronger on complex, open-ended work.
All figures are Anthropic's own.
Does your AI business case track the cost per token, or the cost per finished task?
Image: AI-generated.
Sources
Our file on Anthropic
- 7 Oct
Analysts said Claude subscribers use 42% of Anthropic's computing. They pay 10% of sales.
- 6 Oct
Microsoft planned $1 billion for employees using Claude. Then it cut over a third.
- 3 Oct
Epic stopped building new products for six weeks. An AI model found the flaws.
- 1 Oct
Anthropic said robots could do most physical tasks. They're cost-competitive on 0.3%.
- 1 Oct
Trump ordered federal agencies to rename AI. The new word is Super Intelligence.
Every story here is open to read. The ERP LEADERS brief goes one step further.
The week in enterprise software, in one email: the stories that mattered, the failures with their figures, one case taken apart. Read a sample or sign up for the brief.
Welcome back. · Issue 01
