Anthropic's Claude Sonnet 5 rivals Opus at 60% lower cost

Anthropic has released Claude Sonnet 5, calling it the most agentic Sonnet model yet. The model narrows the gap between mid-tier and flagship Opus-class models, performing nearly as well as Opus 4.8 on agentic tasks like reasoning, tool use, coding, and knowledge work — but at a fraction of the cost. During an introductory period ending August 31, 2026, pricing is set at $2 per million input tokens and $10 per million output tokens; from September onward, it will rise to $3 and $15 respectively. The model is available across all plans (Free, Pro, Max, Team, Enterprise), in Claude Code, and on the Claude Platform.

Early access customers report that Sonnet 5 successfully completes multi-step workflows that earlier versions often left unfinished, and it can independently verify its own work without explicit instruction. Anthropic also highlighted improved safety: the model refuses malicious requests more reliably, resists prompt injection, and reduces hallucinations and sycophantic responses compared to Sonnet 4.6. While it can perform routine cybersecurity tasks, it was not trained for offensive exploits and scores lower than Opus models in that area. Built-in safeguards detect and block dangerous cyber-related requests in real time.