Anthropic released Claude Sonnet 4.6 today. Sonnet is the middle model in Anthropic's lineup — the one most users get by default, without choosing anything. The company describes the release as "a full upgrade of the model's skills across coding, computer use, long-reasoning, agent planning, knowledge work, and design." Pricing is unchanged from Sonnet 4.5, starting at $3 per million input tokens and $15 per million output.

The comparisons on the launch page are to Anthropic's own flagship, Opus. "Performance that would have previously required reaching for an Opus-class model is now available with Sonnet 4.6," the company says. It also says Sonnet 4.6 "matches Opus 4.6 performance on OfficeQA" and "has meaningfully closed the gap with Opus on bug detection." Per Anthropic, "Sonnet 4.6 features a 1M token context window in beta" — enough room to keep very long documents and histories in a single working session.

Availability is immediate and broad: Sonnet 4.6 is now the default model in claude.ai and Claude Cowork, and it ships across all Claude plans, Claude Code, the API, and all major clouds. Axios's headline that day: "Anthropic's basic model is almost as smart as its advanced model."

The gap between tiers is measured in months

That Axios headline is the real story. Opus 4.6 was last quarter's reason to pay for the top tier. Today, on Anthropic's own numbers, the default model matches it on office work — at the price the middle tier already cost. Nobody has to order anything, migrate anything, or approve a new line item. The model under millions of existing accounts got quietly better at the flagship's old job.

Look at what Anthropic chose to highlight. The release says the model is especially strong on "branched and multi-step tasks like contract routing, conditional template selection, and CRM coordination." That is not a research-lab demo list. That is the back office of a 10-person services company: which agreement goes to whom for signature, which proposal template fits this job, whether the CRM record matches what actually happened on site. Frontier models used to be marketed on PhD-level exams; this one is marketed on the work your office manager does every Tuesday.

The pattern worth planning around: capability now trickles down a tier in months, not years. What required the expensive model in the fall runs on the default one by spring, at prices that didn't move. And this release isn't a one-off — the same trickle-down is the whole plot of Claude Sonnet 5, the next chapter in this story.

"Wait for it to get cheaper" is still a rational strategy. It's just a much shorter wait than most owners assume — one or two quarters, not a product generation. That changes what waiting costs. The model improves on its own schedule whether you act or not; the thing that doesn't exist until you build it is a scoped, tested workflow with your documents, your templates, and your CRM plugged in. The scoping is the durable asset — every model release after that is a free upgrade to it. Matching that effort to the task is how we run every engagement.

What to do with this

  • Re-run your parked list. If an automation idea was shelved because the capable model cost too much, the default tier may now cover it at $3/$15. Put it back through the three-question filter before re-funding it.
  • Start with the workflows Anthropic named. Contract routing, template selection, CRM coordination — count the hours per week each one takes in your business, and scope the biggest number first.
  • Check what you already have. If your team uses claude.ai on any plan, they got Sonnet 4.6 today as the default. Testing it against one real workflow costs an afternoon, not a purchase order.
  • Scope now instead of waiting for the next release. The next model will arrive on schedule; your process documentation won't write itself. A 30-day first project is sized so the trickle-down works for you instead of ahead of you.

If you want a second opinion on which of your workflows this release actually reaches, that's a 30-minute conversation.

Source: Introducing Sonnet 4.6 — Anthropic, February 17, 2026