Daily AI CatchupClaude Sonnet 5 Debuts as Anthropic's Most Agentic Sonnet Yet
Anthropic's Sonnet 5 launches as the first model in its new "5" tier, delivering near-Opus 4.8 capability in coding, planning, and tool use at $2/$10 per million tokens through August 31. It can operate a browser or terminal directly and sustain longer multi-step jobs — though cybersecurity benchmarks dipped versus Sonnet 4.6, a byproduct of Anthropic not training on those tasks during the export dispute.
Why it matters
💻 Developer · Near-Opus coding and tool-use capability at Sonnet pricing is worth a head-to-head benchmark against whatever model currently anchors your agentic workflows.
📦 Product · A cheaper model hitting your premium tier's capability bar is a pricing-tier reshuffle waiting to happen in your own product.
🎨 Design · Longer sustained agentic runs mean fewer check-ins needed for multi-step creative or content tasks.
📈 Business · Introductory pricing runs through August 31 before jumping from $2/$10 to $3/$15 per million tokens — worth locking in volume now if you're evaluating a switch.
🤔 Just Curious · Anthropic's new mid-priced model performs almost as well as its most expensive one — a meaningful jump in how much capability your money buys.
Try this: Benchmark Sonnet 5 against your current production model on a real agentic task before the August 31 price increase.
Sources: Anthropic: Claude Sonnet 5