On July 24, 2026, Anthropic announced Claude Opus 5 — and the announcement became one of the clearest signals yet that the "most powerful model" race is no longer enough on its own. Opus 5's pitch is simple: performance nearly matching Fable 5, at half the price.
Within 0.5 Points of Fable 5
According to Anthropic's own data, Opus 5 lands within 0.5 points of Fable 5's peak score on the CursorBench 3.2 benchmark when run at max effort. In practice, that means most users won't notice much difference between the two models in daily work — but the price tag is a different story entirely.
Anthropic also claims Opus 5 beats every other model at the same cost point on CursorBench when run at high, extra-high (xhigh), and max effort settings.
Looking closer at the numbers: Opus 5 scores 96.0% on SWE-bench Verified, 79.2% on SWE-bench Pro, and 43.3% on Frontier-Bench. For comparison, the previous-generation Opus 4.8 scored 88.6% on SWE-bench Verified and 69.2% on SWE-bench Pro — meaning Opus 5 posted a double-digit jump over its own predecessor. On SWE-bench Pro specifically, Opus 5 lands third within the family, trailing Mythos 5's 80.3% and Fable 5's 80.0% by under a point. In other words, Opus 5 isn't pitched as the outright leader — it's pitched as landing within a hair of the top while cutting the price in half.
A Five-Tier "Effort" System
Opus 5's standout feature is letting users pick between five distinct effort levels for every single request:
- Low
- Medium
- High
- Xhigh
- Max
This system lets teams deliberately dial down reasoning depth on simple tasks to save tokens and latency, then switch to full power for complex ones. In practice, that means getting the flexibility of both a cheap-and-fast model and an expensive-and-powerful one from a single subscription.
Pricing and Savings
Opus 5 is priced at $5 per million input tokens and $25 per million output tokens — the exact same price point as the previous-generation Opus 4.8, but with noticeably stronger performance. On top of that:
- Prompt caching can cut costs by up to 90%
- Batch processing can cut costs by up to 50%
Combined, these two features can meaningfully reduce total costs for high-volume enterprise workflows.
The New Default for Claude Max
Anthropic is positioning Opus 5 as the "everyday model" for enterprises, knowledge workers, and developers. It became the default model for Claude Max subscribers — a clear sign that Anthropic now sees the Opus line not as a "luxury" tier, but as an accessible standard for a broad user base.
Why This Matters
Claude Opus 5's launch captures one of the clearest trends of summer 2026: price-to-performance now matters more than raw performance alone. The same week, Grok 4.5's aggressive pricing and GPT-5.6's three-tier structure carried a similar message — users no longer just want "the best model," they want the best model for the job at the lowest reasonable cost.
Opus 5 was a well-considered move for Anthropic in that race, strong on both the technical and commercial fronts.
