Claude Sonnet 5.5: Anthropic's second new model in a week

What's new

Anthropic, which is backed by Google among others, has released two new AI models in less than a week. First came Claude Opus 5.5 on September 22, and now Claude Sonnet 5.5 follows on Monday, September 28, 2026. Sonnet 5.5 is the middle model in the Claude 5.5 family and is meant to serve as a faster and cheaper alternative to the most powerful model, Opus 5.5. A third and cheaper model, Haiku 5.5, is on the way in the coming weeks. The pace is remarkable because Anthropic CEO Dario Amodei just two weeks ago urged the entire AI industry to slow down development for the sake of safety.

What's clever

Sonnet 5.5 responds more than 30 percent faster than its predecessor, Sonnet 5, and typically uses fewer "tokens" (the pieces of text an AI model splits language into) to solve the same task. In practice, that makes it cheaper to use, even though the price per token itself is unchanged. The model can, among other things, fix bugs in code, write documents and create slides and spreadsheets. Anthropic also says the safety level matches the far more powerful Opus 5 model, and that Sonnet 5.5 is the least inclined of all tested Claude models to look for security holes on its own.

Cheaper or better?

The price per token for Sonnet 5.5 is unchanged from Sonnet 5: $2 per million input tokens and $10 per million output tokens (about DKK 14 and DKK 70). Because the model solves tasks faster and with fewer tokens, Anthropic says the total price for a typical task drops by up to 30 percent. Big brother Opus 5.5 even got a direct price cut last week: input fell from $5 to $4 per million tokens, output from $25 to $20, and cache reads became 60 percent cheaper. Overall, Opus 5.5 runs about 40 percent cheaper than its predecessor. Regular users don't notice the price itself, as Sonnet 5.5 is free to use via Claude.ai on the web, iPhone and Android. It is mainly companies and developers who pay per token via Anthropic's developer platform, AWS, Google Cloud or Microsoft Foundry who feel the savings directly.

What it's good at

Sonnet 5.5 is tailored to the tasks most employees deal with every day: debugging code, writing documents, setting up slides and spreadsheets. According to Anthropic, customers such as Slack and Unity solve tasks in fewer steps and with significantly fewer tokens than before. If, on the other hand, you need AI for the hardest and most complex problems, the more expensive Opus 5.5 is still the right model to reach for.