Anthropic's New AI Is 30% Faster and Cheaper for Everyday Work

Anthropic, the company behind the Claude chatbot, just launched a new version of its middle-of-the-road AI called Sonnet 5.5. Think of Anthropic's models like cars: Opus is the powerful, expensive flagship, Haiku is the tiny economy model, and Sonnet is the sensible mid-size sedan most people actually drive. This new Sonnet is the sedan with a better engine. Anthropic says it runs 30% faster than the version released just three months ago, and it uses far fewer "tokens" — the units AI consumes to read and write, roughly like minutes on a phone plan. Fewer tokens means lower bills for anyone paying per use.

The bigger deal is what it enables. Sonnet 5.5 is designed to run "agents" — AI that can take actions on your behalf, like editing files, filling in spreadsheets, or writing and testing code without you babysitting it. Anthropic says Sonnet 5.5 actually outperforms its pricier Opus model on coding tasks, mainly because it can run several agents at once without the cost spiraling. For a small business, that's the difference between automating a task and deciding it's not worth the money.

There's a security angle too. Anthropic claims Sonnet 5.5 has cyber abilities on par with Opus 5, meaning it could be used to find or exploit software flaws. So for the first time, a Sonnet model gets the same extra safety guardrails Anthropic applies to its most powerful systems. In practice, that means the AI may refuse some requests or require extra checks — a trade-off for safety.

The timing matters. AI releases are arriving almost weekly: OpenAI shipped new models last week, Meta unveiled one for its smart glasses, and Anthropic says a new, smaller Haiku is coming in weeks. The pattern is clear — AI is getting faster and cheaper fast. Whatever you pay for AI help today will likely buy more tomorrow.