News

Claude Sonnet 5 Is Here — And It's Fully Agentic

Editorial Team · 2026-07-02 · 7 min read

Anthropic's newest release delivers near-Opus 4.8 performance at Sonnet pricing — and it can autonomously use browsers, terminals, and tools. We tested it

Anthropic has released Claude Sonnet 5, its most agentic model yet. The launch coincides with the US government lifting national security restrictions that had briefly pulled the Fable 5 and Mythos 5 models offline for three weeks in June.

Sonnet 5's headline feature is autonomous tool use. Unlike previous models that required careful prompt engineering to chain actions, Sonnet 5 can independently decide to open a browser, run a shell command, query an API, and synthesize the results — all without human intervention between steps.

In our testing, this changes the economics of AI agents dramatically. Tasks that previously required Opus (at 5x the cost) now run reliably on Sonnet 5. The gap between the flagship and the workhorse model has essentially closed for agentic workloads.

What this means: if you're building AI agents, you can now deploy them at scale without watching your token bill explode.