Anthropic shipped Claude Sonnet 5 on June 30 and made it the default on Free and Pro. The pitch is simple. You get performance close to Opus 4.8 on agents, coding, and tool use, at a fraction of the price. For anyone running agent loops in production, that is the number that matters.
The pricing that moves the needle
Through August 31, Sonnet 5 runs at $2 per million input tokens and $10 per million output. After that it climbs to $3 and $15. Either way it undercuts Opus 4.8, GPT-5.5, and Gemini 3.1 Pro. When your workload is thousands of tool calls a day, that gap compounds fast.
The catch nobody is pricing in
Sonnet 5 ships with a new tokenizer, and the same text can map to up to 1.35 times more tokens than before. Anthropic set the intro price to keep the switch roughly cost neutral, but that cushion expires with the promo. Before you migrate an agent, run your real traffic through it and measure tokens per task, not just latency. That number tells you whether Sonnet 5 is a genuine win or just a cheaper looking bill.