Anthropic released Claude Sonnet 5 on Tuesday and made it the default model for free and Pro plans, putting a more powerful agentic version of its midsize system into everyday use. The company says the model can make plans, use tools such as browsers and terminals, and run on its own at a level that only larger, pricier models could handle a few months ago.
The timing matters because buyers are now shopping for models that can do real work, not just answer prompts. Claude Sonnet 5 arrived at $2 per million input tokens and $10 per million output tokens through August 31, then moves to $3 per million input tokens while the output price stays flat. That makes it cheaper than Opus 4.8 and some other high-end options, but still more expensive than Gemini 3.5 Flash, which leaves Anthropic trying to sell value rather than the absolute lowest price.
Anthropic is also arguing that the new model closes much of the gap with its top tier. It said Claude Sonnet 5 delivers significant gains over Sonnet 4.6 in reasoning, tool use, software coding and knowledge work, while promising performance close to Opus 4.8 at far lower cost. On agentic coding, Sonnet 5 scored 63.2%, ahead of Sonnet 4.6 at 58.1% but behind Opus 4.8 at 69.2%. On a knowledge work benchmark, it slightly beat Opus 4.8, even as Anthropic said Opus 4.8 remains the better choice when accuracy is the priority.
That gap is exactly what makes the release interesting. Anthropic wants developers to see a model that is good enough for autonomous workflows without paying top-tier rates, and it is backing that pitch with a practical claim: testers said Claude Sonnet 5 finishes complex jobs where earlier versions would have stopped short and checks its own output without being asked.
Daniel Shepard, who described one of those tests, said Claude Sonnet 5 handled a two-part automation job that had previously broken down midway. He said it updated Salesforce account tiers and sent a launch announcement to enterprise contacts in one run. “That used to stall halfway. For day-to-day automation, it’s a no-brainer,” he said.
The launch also reflects how quickly agentic features have become the baseline in the model market. Anthropic says the system is safer to use in agentic settings and shows a lower rate of undesirable behavior such as cooperation with misuse and deception than its predecessor. OpenAI’s GPT-5.6 Sol appeared in preview last week, Google’s Gemini 3.5 Flash arrived in May, and Anthropic is now trying to meet that competition with a model meant to sit in the middle: more capable than Sonnet 4.6, cheaper than Opus 4.8, and priced to win the kind of work that can be automated at scale. The key question after Tuesday is how quickly developers adopt it before the August 31 price change narrows the window of its launch pricing.
