Anthropic's Claude Sonnet 5 Beats Opus-Class AI at 40% Lower Cost

Anthropic has released Claude Sonnet 5, the most capable model in its Sonnet family and a significant upgrade over Sonnet 4.6. The pitch is simple: near-Opus intelligence at Sonnet pricing, built specifically for agents that plan, use tools, and run autonomously on complex multi-step work. "It can make plans, use tools like browsers and terminals, and run autonomously at a level that, just a few months ago, required larger and more expensive models," Anthropic said at launch.

Claude Sonnet 5 is now the default model for free and Pro plans and is available for every subscription. It's accessible through the Anthropic API, Claude.ai, Amazon Bedrock, and Google Cloud Vertex AI.

The numbers that matter

At launch, Sonnet 5 is priced at $2 per million input tokens and $10 per million output tokens through August 31, after which the price will jump to $3 per million input tokens and $15 per million output tokens. That post-August price is identical to what Sonnet 4.6 cost, so the rate card looks flat , but there is a catch. The model also carries a new tokenizer that can turn the same text into up to 35% more billable tokens. Teams migrating from Sonnet 4.6 should measure their actual token delta before treating this as a cost-neutral swap.

Sonnet 5 performs close to Opus 4.8 on agentic work at a fraction of the price. Opus 4.8 is $5/$25 per million tokens. Sonnet 5 is 40% cheaper on both at standard rates, and 60% cheaper during the intro window.

What AA-Briefcase reveals

The most interesting data comes from AA-Briefcase, a new proprietary benchmark from Artificial Analysis that tests agentic knowledge work. It runs models through realistic multi-week professional projects , data science, product management, banking operations, corporate strategy , grading deliverables like spreadsheets, presentations, and memos across three dimensions: binary rubric checks for factual correctness, pairwise analytical quality, and pairwise presentation quality.