Anthropic has unveiled Claude Sonnet 5, a new midsize AI model designed to deliver stronger agentic performance at a more accessible price point. The company says the model can plan tasks, work with tools such as browsers and terminals, and operate with less human input than earlier versions.
The launch reflects a wider shift across the AI industry, where agentic systems are becoming the new standard. Recent releases from OpenAI and Google have also emphasized models that can break down work, coordinate tasks, and complete longer workflows with minimal supervision.
Anthropic positions Sonnet 5 as a practical balance between capability and cost. The model is set to become the default option for free and Pro users, while remaining available across all subscription tiers. It is priced below Anthropic's higher-end offerings and under several competing models, though it is still above some lightweight alternatives.
According to Anthropic, Sonnet 5 improves on Sonnet 4.6 in reasoning, coding, tool use, and knowledge work. In internal benchmarks, it narrows the gap with the company's more advanced models and even edges ahead in some knowledge-based tasks. Testers also reported that the model is better at completing complex assignments end to end and can review its own output more independently.
Safety has also been a focus. Anthropic says Sonnet 5 shows fewer problematic behaviors, including lower rates of deception, misuse cooperation, and prompt-injection susceptibility. The company notes that it is designed to refuse unsafe requests more consistently, making it better suited for agentic environments.
With Claude Sonnet 5, Anthropic is signaling that the next phase of AI will be defined not only by power, but by efficiency, reliability, and responsible autonomy. That balance could shape the future of everyday digital work.
