Claude Sonnet 5 brings near-flagship AI agents down to a mid-tier price

Anthropic closed out June with a release aimed squarely at people who found its best models too expensive to run all day. Claude Sonnet 5, launched June 30, is the company's new mid-tier model, and its pitch is simple: most of the ability of a flagship, at a fraction of the price.

What Sonnet 5 can do

The focus this generation is agentic work — AI that does not just answer questions but makes a plan, uses tools like a web browser or a terminal, and carries a task through to the end on its own. On Anthropic's agentic coding benchmark, Sonnet 5 scores 63.2%, up sharply from the previous Sonnet 4.6 at 58.1% and within striking distance of the far pricier Opus 4.8 at 69.2%. On general knowledge work, it actually edges past Opus 4.8.

Early users report the difference is less about individual answers and more about follow-through: multi-part tasks that older models abandoned halfway now get finished, and the model checks its own output without being told to. Anthropic also says Sonnet 5 shows lower rates of hallucination, deception, and tell-you-what-you-want-to-hear behavior than its predecessor.

The pricing play

Sonnet 5 launches at $2 per million input tokens and $10 per million output tokens through August 31, after which it moves to $3 and $15. Either way, it undercuts Opus 4.8, GPT-5.5, and Gemini 3.1 Pro. It is also now the default model for Claude's free and Pro plans, so most Claude users get the upgrade automatically.

The economics matter because agents are token-hungry. A model that works autonomously for twenty minutes burns through far more tokens than a quick chat, so the per-token price largely determines whether automating a workflow is profitable or a novelty.

A good week for Anthropic overall

The launch capped an eventful month: the same week, the US government lifted the export-control order on Anthropic's Fable 5 and Mythos 5 models, and Fable 5 returned to service on July 1 after a 19-day shutdown. Anthropic now has both its frontier tier and a newly capable budget tier back on the table.

Why it matters for small business

This is the release that makes AI agents affordable for smaller operations. Tasks that were technically possible but too costly to automate — monitoring competitors weekly, reconciling orders against invoices, drafting and organizing routine documents end to end — now run on a model priced for volume. If you use Claude's free or Pro plan, you already have Sonnet 5; try handing it one complete multi-step chore rather than a series of one-off questions. And if you pay a vendor for an AI-powered service built on Anthropic's models, their costs just dropped — a fair topic at your next renewal conversation.