Anthropic launches Claude Sonnet 5

Anthropic has launched Claude Sonnet 5, its new default model for everyday Claude users and a more capable option for developers building agents. In a June 30 product post, the company says Sonnet 5 is its most agentic Sonnet model yet: it can plan, use tools such as browsers and terminals, and work autonomously on tasks that recently required larger, more expensive models. Anthropic positions the release as a cost-performance bridge between Sonnet 4.6 and Opus 4.8.

The availability is broad from day one. Sonnet 5 is now the default model for Claude Free and Pro users, and it is available to Max, Team and Enterprise customers. It is also available in Claude Code and through the Claude API as claude-sonnet-5. On the platform, Anthropic is launching it at an introductory price of $2 per million input tokens and $10 per million output tokens through August 31, 2026. After that, the standard price becomes $3 and $15 per million tokens.

Why does this matter? Sonnet is the model tier many users actually touch every day. Opus may carry the prestige of maximum capability, but Sonnet often defines the practical experience: coding help, research, document work, automation and tools inside products. If Sonnet 5 gets close to Opus 4.8 on some agentic tasks at a lower price, more teams can move from prompt-and-response workflows to agents that plan, call tools, test output and finish multi-step jobs.

Anthropic is also trying to frame the release around safety. The company says Sonnet 5 has lower rates of undesirable behavior than Sonnet 4.6, is safer in agentic contexts, and performs far below Opus and Mythos models on dangerous cyber evaluations. It still ships with cyber safeguards enabled by default.

For AI users, the short-term change is simple: Claude's default experience should become more capable. For makers and businesses, the bigger shift is economic. Better mid-tier models lower the cost of embedding agentic AI into products, internal tools and software engineering workflows. The competition is no longer only about the strongest frontier model; it is about which lab can make autonomy cheap enough, safe enough and reliable enough for daily work.