Anthropic has released Claude Sonnet 5, positioned as its most agentic mid-tier model yet. Its performance lands close to the flagship Opus 4.8 on reasoning, coding, and tool use — at a noticeably lower price. Here's how it performs across the three areas that matter most.
1. Agentic Coding
Agentic coding means writing code with full command of the entire codebase. Here Sonnet 5 sits right between Sonnet 4.6 and Opus 4.8, scoring 63.2% and beating many competing models from other companies — though it still trails Opus 4.8 (69.2%) by roughly six points.
The biggest leap shows up in terminal control and operating inside terminal environments. The previous model, Sonnet 4.6, scored 67% here, while Sonnet 5 comes within about 2 points of Opus 4.8 and delivers around 80% consistency.
2. Multidisciplinary Reasoning
This category tests PhD-level problem solving, and it's another area where Sonnet 5 makes a jump. Without tools, Opus 4.8 keeps a ~5-point edge at 50%. But once tools are enabled, Sonnet 5 catches Opus and reaches 57%.
3. Computer Use
When it comes to using a computer the way we do, there's little separating Sonnet 5 from Opus 4.8. Sonnet 5 scores 81.2%, leaving many other models behind.
Pricing
As an introductory rate through August 31, Sonnet 5 runs $2 per million input tokens / $10 per million output tokens, then rises to $3 / $15. For this level of capability, it's a genuinely affordable model.
So — what do you think about it?