Anthropic launches Claude Sonnet 5 to bring near‑Opus agents at lower cost

Sonnet 5 is strategically important because it shifts Anthropic’s strongest agentic capabilities from the ultra‑premium tier into the mass‑market middle tier. By making a model that approaches Opus 4.8 on coding, tool use, and long‑running agent tasks the default for free and Pro users, Anthropic is raising the baseline of what everyday “assistant” models can do. The Japanese and Argentine coverage underscores a core message: Sonnet 5 is designed to deliver Opus‑like autonomy at meaningfully lower prices, while keeping more dangerous cyber capabilities reserved for higher‑risk, more tightly controlled systems.

For the race to AGI, the more interesting angle is distribution, not raw benchmarks. Sonnet 5 ships everywhere Claude runs and integrates quickly into Amazon Bedrock and the Claude Platform on AWS, making it trivial for enterprises to swap in a more capable agentic core without re‑architecting their stacks. That accelerates the practical deployment of multi‑step agents across software engineering, back‑office workflows, and knowledge work. At the same time, Anthropic is explicitly framing Sonnet 5 as a non‑frontier model under its scaling policy, signaling a deliberate attempt to grow agentic usage without escalating capability risk. In effect, Sonnet 5 turns what used to be a frontier capability—autonomous, tool‑using agents—into something that can be treated as mid‑tier infrastructure.