Anthropic just dropped Sonnet 5, and it’s surprisingly good. The company says it’s the most agentic Sonnet model yet — capable of planning, using tools, browsing the web, and running terminal commands. Features that were mostly locked to Opus 4.8 are now available at a lower tier.
For developers, this is a big deal. Sonnet 4.6 wasn’t great at handling complex multi-step tasks or understanding large codebases. Sonnet 5 changes that. Testers described it as “much more agentic than its predecessors” — it can check its own output without being asked and handles “can you fix your code?” queries far better.
Benchmarks back this up. BrowseComp and OSWorld-Verified scores put Sonnet 5 significantly closer to Opus 4.8 than any previous Sonnet model.
Pricing is where it gets interesting. Sonnet 5 launches at $2 per million input tokens and $10 per million output tokens through August 31. After that, it bumps to $3 and $15 respectively. Compare that to Opus 4.8 at $5 and $25. You’re getting near-flagship performance for roughly 40% less.
Regular users don’t need to worry about pricing tiers. Sonnet 5 is available across Free, Pro ($20/month), and Max ($200/month) subscriptions.
Anthropic’s strategy is clear: don’t lock agentic gains behind the most expensive models. Sonnet 5 brings advanced capabilities to a broader audience. And for anyone who’s been burning through Opus tokens on everyday coding tasks, this is exactly what they’ve been waiting for.
