Anthropic has launched its Claude Sonnet 5 model in a new step within the race of intelligent agents, where the goal is no longer just to provide quick text answers but also to perform extended tasks such as planning, using browsers, handling tools, writing code, and following through with work until completion. The company is betting on a less expensive model than its larger ones, yet capable of performing tasks that previously required heavier and more expensive models. This places it directly in the heart of competition for the practical artificial intelligence market within companies.

An intelligent tool-driven agent

According to an announcement published on Anthropic’s website, Claude Sonnet 5 is now available across all plans. It is the default model for Free and Pro users, and also available for Max, Team, and Enterprise users. It operates through Cloud Code and the Cloud Developer Platform. The company states that its introductory price is $2 per million input tokens and $10 per million output tokens until August 31, 2026, after which it will move to $3 and $15.

The cheaper model gets closer to the stronger ones

Anthropic says that Sonnet 5 performs close to Ops 4.8 in some tasks, particularly in proxy search and computer use. However, it gives users a wider range of choices for balancing price and performance. Anthropic also notes that the model has shown fewer undesirable behaviors than Sonnet 4.6 and less capability for performing cybersecurity tasks compared to Ops models. This is an important detail at a time when concerns are increasing about models capable of independent work within browsers and systems.