Anthropic just cannibalized its own flagship model — and that’s exactly the point.

Claude Sonnet 4.6, released today, delivers near-Opus performance at $3/$15 per million tokens. That’s one-fifth the cost of the model it’s chasing. In benchmark after benchmark, the mid-tier model is breathing down its big brother’s neck. In some cases, it’s already ahead.

This isn’t a minor version bump. It’s a price-performance earthquake.

Two Models, Twelve Days, One Message

Anthropic dropped Opus 4.6 on February 5th with “Agent Teams” — the ability to spin up multiple AI agents that coordinate in parallel. Think less “chatbot” and more “autonomous project team.” A million-token context window. PowerPoint integration. The works.

Twelve days later, Sonnet 4.6 arrives and nearly matches it across the board.

The message is clear: frontier AI capability is commoditizing faster than anyone expected.

The Numbers That Matter

Forget the benchmarks that only matter in papers. Here’s what developers actually care about:

  • SWE-bench Verified: 79.6% — frontier territory
  • Computer use accuracy: 94% on the Pace insurance benchmark, the highest of any model tested
  • OSWorld: 72.5%, just 0.2% behind Opus 4.6
  • Developer preference over Sonnet 4.5: 70% in Claude Code testing
  • Developer preference over Opus 4.5: 59% — a mid-tier model beating last gen’s flagship

Read that last one again. Most developers preferred the cheaper model over the previous top-of-the-line. That’s not incremental progress. That’s a generational leap compressed into months.

What Developers Are Actually Saying

The benchmark story is compelling. The real-world feedback is even better.

Cursor’s co-founder called it “a notable improvement across the board.” Mercury Banking said the speed-cost-quality trifecta was “a surprising combination of improvements.” Cognition — the team behind Devin — found it “meaningfully closed the gap with Opus on bug detection,” letting them run more parallel code reviewers without blowing their budget.

But the qualitative improvements might matter more than the quantitative ones. Developers report Sonnet 4.6 actually reads your code before modifying it. It picks up project conventions. Ask for a simple fix, you get a simple fix — not an unsolicited refactor of your entire module.

If you’ve ever rage-quit an AI coding session because the model rewrote half your codebase to fix a typo, this is the update you’ve been waiting for.

Opus 4.6: The Agent Teams Era

Don’t sleep on Opus 4.6 just because Sonnet is stealing the spotlight.

Agent Teams is the real paradigm shift. Instead of one AI grinding through tasks sequentially, Opus 4.6 splits work across multiple coordinated agents running in parallel. Anthropic’s Head of Product compared it to “having a talented team of humans working for you.”

Combined with the million-token context window, this is where AI stops being a tool and starts being a workforce. Feed it an entire codebase. Let the agents divide and conquer. This is what “vibe working” looks like in practice.

Pricing sits at $5/$25 per million tokens — reasonable for a model that’s essentially replacing junior dev teams.

A Trillion-Dollar Wake-Up Call

Wall Street noticed. The iShares Expanded Tech-Software Sector ETF plunged over 20% year-to-date after the Opus 4.6 release. A trillion dollars evaporated from software stocks.

The logic is brutal but simple: if AI can automate complex knowledge work — legal analysis, financial modeling, software engineering — at $3 per million tokens, what happens to the companies charging humans to do it?

Anthropic’s $380 billion valuation from its recent $30 billion Series G says the market has picked a winner. Or at least a category.

Sonnet 5 Is Already Lurking

Here’s the kicker: a “claude-sonnet-5@20260203” identifier was spotted in Google Vertex AI error logs. Multiple independent sources corroborated it.

If Sonnet 4.6 already matches Opus 4.6 on several benchmarks, a true next-generation Sonnet could reset the entire competitive landscape. Again.

What This Means For You

If you’re a developer: Switch from Opus 4.5 to Sonnet 4.6. You’ll save roughly 60% and probably get better results. The improved context comprehension and reduced overengineering make it a no-brainer for daily workflows. The 1M context window beta means entire repos are fair game.

If you’re running a business: Sonnet 4.6 is now the default free-tier model. You can prototype with frontier AI at zero cost. Anthropic’s new Infosys partnership is already building enterprise agents for banking, telecoms, and manufacturing. The gap between “works in a demo” and “works in production” is closing fast.

If you’re in software: The disruption clock is ticking louder. What was flagship-tier performance two months ago is now free. The question isn’t whether AI transforms knowledge work. It’s whether you’ll be riding the wave or under it.

The AI price-performance curve isn’t just steep — it’s vertical. Buckle up.