Anthropic released Claude Sonnet 5 today — the middle model in its lineup, and the one most businesses will actually end up using. The company calls it "our most agentic Sonnet yet, with top-tier intelligence for coding and everyday professional work."

Two facts from the announcement matter more than the rest.

The headline: near-flagship results at mid-tier prices

Anthropic says Sonnet 5's performance is "close to that of Opus 4.8" — its flagship — "but at lower prices," and in some cases matches the flagship's capability outright. Pricing starts at $2 per million input tokens and $10 per million output tokens through August 31, moving to $3 and $15 after that.

If token pricing means nothing to you, here's the translation: the work most businesses want automated — drafting replies, summarizing documents, turning notes into quotes, keeping records consistent — costs pennies per task at these rates. The model is rarely the expensive part of an AI project anymore. The expensive part is choosing the wrong project, which is a picking problem, not a technology problem.

The quieter story: built for sustained work

The capability Anthropic emphasizes isn't a benchmark score — it's endurance. Sonnet 5 is pitched at "handling sustained coding, tool use, and debugging well across messy technical contexts." That phrase, stripped of the engineering framing, describes most real back-office work: multi-step, interrupted, dependent on several systems, and full of exceptions.

Earlier models were strong at single answers and shaky at step twelve of a forty-step task. The current generation — and Sonnet 5 is squarely in it — is being trained specifically to stay coherent across long, tool-heavy sessions. That's the difference between AI that answers questions about your invoice process and AI that works your invoice process.

One more line worth noticing for anyone nervous about reliability: Anthropic reports Sonnet 5 scored safer than its predecessor across its evaluations for undesirable behaviors. Vendor-reported, like all launch claims — but it signals where the effort is going.

What to do with this

  • If you already run AI workflows: re-check your model choice. Sonnet 5 is the default on Anthropic's Free and Pro plans and available via the API — if you're paying flagship prices for work a mid-tier model now handles, that's margin you're leaving on the table.
  • If you've been waiting because "it's too expensive": that reason is mostly gone. The intro pricing window (through August 31) is a low-risk moment to run a 30-day first project.
  • If you're not sure where it fits: that's the actual question, and it hasn't changed with the release. Model launches change what's affordable, not what's worth doing — finding the right use case is still the work.

We read the release notes so you don't have to. If you'd rather talk through what this means for your specific operation, the first conversation is 30 minutes and pitch-free.

Source: Introducing Claude Sonnet 5 — Anthropic, June 30, 2026