Anthropic's New Claude Sonnet 5 Is Cheap, Fast, and Built to Act
Anthropic is capping a big week with a new model. Claude Sonnet 5 launches with performance approaching its flagship Opus at a fraction of the price, roughly 2 dollars in and 10 dollars out per million tokens, and Anthropic is calling it its most agentic Sonnet yet. It is a bet that cheap, capable, action-taking AI is what businesses actually want.
Anthropic is capping a big week with a new model. Claude Sonnet 5 launches with performance approaching its flagship Opus at a fraction of the price, roughly 2 dollars in and 10 dollars out per million tokens, and Anthropic is calling it its most agentic Sonnet yet. It is a bet that cheap, capable, action-taking AI is what businesses actually want.
The pitch is power without the price. Sonnet is Anthropic's mid-tier model, and getting it close to top-tier quality while keeping the cost low is exactly what most companies need, since they run AI millions of times and cannot afford the priciest model for every task. Calling it the most agentic Sonnet means it is tuned to take actions and complete tasks, not just answer questions. Useful and affordable is the combination.
It fits a clear shift in the market. Users have moved from wanting the biggest, most expensive model for everything toward efficiency, choosing models that are good enough and cheap enough to deploy at scale. The whole industry is racing to deliver capability at lower cost, and a strong mid-tier model is how a lab wins the high-volume business that pays the bills. Efficiency is the new battleground.
Anthropic is pairing it with distribution wins. Claude is now generally available on Microsoft's Azure AI platform, running on Nvidia's newest chips, and California struck a deal to bring Claude to all its state agencies, cities and counties at a steep discount. Getting into a major cloud and a giant government customer matters as much as the model itself. Reach turns a good model into a business.
The honest caveat is that everyone is doing this. Google just pushed cheap, fast models, and rivals are all racing to the same efficient middle, so a strong, affordable model is table stakes rather than a lasting edge. Benchmarks also do not always match real-world use, and today's price leader can be undercut tomorrow. Being cheap and capable is necessary, not sufficient.
So Anthropic is meeting the market where the money is, with a cheaper, more capable, action-oriented Sonnet and the deals to spread it. A near-flagship model at a mid-tier price, an Azure launch, and a statewide government customer. The AI race is moving to value, not just intelligence. Watch adoption of Sonnet 5 and how rivals answer on price.