Anthropic released Claude Sonnet 5 on June 30, 2026, making it the default model for Free and Pro users and the headline of the Claude lineup. For the first time, a Sonnet-class model outscored the concurrent Opus flagship — scoring 1,618 on GDPval-AA v2 — and it does it at a fraction of the price.
The First Sonnet to Beat Opus
Sonnet 5’s 1,618 GDPval-AA v2 score marks the first time a mid-tier Claude has topped the top tier on a major benchmark. Anthropic says performance is now close to Opus 4.8 on reasoning, tool use, coding, and knowledge work, delivered with a 1-million-token context window.
The agentic gap that opened up in the Opus generation has also narrowed: Sonnet 5 can make plans, use browsers and terminals, and run autonomously at levels that just a few months ago required larger, more expensive models. Early partners describe it finishing complex multi-step jobs where previous Sonnet models stopped short, and checking its own output without being asked.
Pricing
- $2 per million input tokens and $10 per million output tokens — versus $5 and $25 for Opus 4.8
- Default model for Free and Pro plans; available on Max, Team, and Enterprise
- Developers access it via the Claude API as
claude-sonnet-5 - Rate limits raised across Chat, Cowork, Claude Code, and the Claude Platform
Safety First
Safety evaluations found Sonnet 5 shows a lower rate of undesirable behaviors than Sonnet 4.6, including reduced hallucination and sycophancy and better refusal of malicious requests. It also has substantially weaker cyber capabilities than Opus-class models — it never developed a working exploit in the Mozilla Firefox evaluation — and ships with real-time cyber safeguards enabled by default.
What This Means
Sonnet 5 resets the cost-performance curve, delivering near-flagship agentic ability at mid-tier prices. With the model now powering the default consumer experience, Anthropic is betting that agentic capability — not raw benchmark size — is what wins mainstream and enterprise users in 2026.