Anthropic unveiled Sonnet 5, a new Claude model optimized for agentic AI tasks that have been straining enterprise budgets. These automated queries generate far more requests than human users, leading to unexpectedly high token consumption and costs. The company designed this model to handle such workloads more efficiently while maintaining performance comparable to recent Opus models.

Starting September 1, Sonnet 5 will cost $3 per million input tokens and $15 per million output tokens, undercutting Opus 4.8's $4 and $25 rates. Early adopters can access even lower pricing before that date. This pricing strategy positions Sonnet 5 as a cost-effective solution for businesses running automated AI agents at scale.

The model launches across Claude Code and the Claude Platform, becoming the default option for free and Pro tier subscribers. Agentic AI tools often exhaust budgets through relentless querying, making efficiency gains essential for sustained operations.

Sonnet 5 incorporates a new tokenizer to squeeze more performance from each token, directly addressing the spending concerns of heavy users. For enterprises managing AI agent fleets, this represents a practical step toward sustainable deployment costs rather than speculative future promises.