Claude Opus 5 Opens a Short Arbitrage Window: Sell Teams a Fixed-Scope Migration From Their Current Model to Whatever Now Wins on Cost or Capability.

Anthropic shipped Claude Opus 5 today, and buried in the announcement is a specific, reproducible business opportunity: for a growing list of tasks, it now matches or beats the more expensive frontier model at a fraction of the price — and most teams paying for the expensive model haven't noticed yet.

Anthropic's own framing is blunt about the tradeoff: Opus 5 is "a thoughtful and proactive model that comes close to the frontier intelligence of Claude Fable 5 at half the price." That's not a vague marketing line — it's attached to numbers you can check.

The gap this creates

On CursorBench 3.2, Opus 5 lands "within 0.5% of Fable 5's peak score at half the cost." On OSWorld 2.0, a benchmark for computer-use agents, it "surpasses Fable 5 at one-third the cost." On ARC-AGI 3, it scores "three times higher than the next-best model." Pricing is unchanged from the previous generation at $5 per million input tokens and $25 per million output tokens, with a fast mode running "2.5x faster" at "2x base cost."

None of that matters to a team unless someone re-runs their specific workflow against the new numbers. Most engineering teams don't re-benchmark their model choice every time a competitor ships — they picked a model months ago and left it running. That gap between "the new best option" and "what's actually in production" is exactly where a service business fits.

Money play

1. Cold-audit a client's current model bill against Opus 5's published numbers before writing any code — the CursorBench and OSWorld deltas alone are enough to justify a paid pilot.

2. Sell a fixed-scope "model swap audit": one recurring, high-volume workflow, benchmarked on the client's own tasks against their current model and against Opus 5.

3. Position the migration as low-risk since pricing didn't change from Opus 4.8 — the buyer is testing a capability upgrade, not absorbing new cost.

4. Prioritize teams currently paying for Fable 5 on tasks where Opus 5 is now "within 0.5%" or ahead at a third to a half of the cost — the easiest sell in the room.

5. Turn the one-time swap into a retainer: benchmark drift is constant in this market, and someone needs to keep watching for the next model that reopens the gap.

Bottom line

The business isn't Opus 5 itself — it's the fact that almost nobody has re-run their model comparison since it shipped. Whoever does that audit first, on a client's real workflow instead of a marketing benchmark, gets paid for closing a gap that already exists in public numbers.

Source: https://www.anthropic.com/news/claude-opus-5