Anthropic Just Made Its Most Powerful AI Model Also the Best Value

Anthropic launched Claude Opus 5 on July 24, 2026 — and the pricing headline hides the real story. The model costs $5 per million input tokens and $25 per million output tokens, the same as Opus 4.8. But on the benchmarks that matter for real work, Opus 5 matches or beats the leading competitors at half the cost. That changes the value equation for anyone using AI seriously.

What actually changed

The price tag stayed flat. The capability did not.

On Frontier-Bench v0.1, Opus 5 more than doubles Opus 4.8’s performance at the same price point.
On CursorBench 3.2 (the closest thing to a real coding test), it lands within 0.5% of Fable 5’s performance at half the cost.
On OSWorld 2.0 (computer-use agents), it surpasses Fable 5’s best result at one-third the cost.
On ARC-AGI 3, it scores three times higher than other leading models.

Why the “half the price” framing matters

The AI pricing war has quietly been running in users’ favor. Until recently, the most powerful models from different labs cost roughly the same. Now the math is shifting: Anthropic is offering the same quality at a meaningfully lower price per useful output. For businesses running agentic workflows, that’s not a rounding error — it’s a budget line.

The model is now the default on Claude Max and particularly strong in agentic coding, complex debugging, root-cause analysis, financial modeling, and visual outputs.

The one-line takeaway

Opus 5’s price didn’t drop — its value went up. If you’re comparing AI tool costs in 2026, the relevant number is cost per task completed, not cost per token. Opus 5 is the first flagship model where that number genuinely puts the competition on notice.

Want the calm version of AI news like this, once a week? Subscribe to the Sharp AI Hub newsletter →

Sources: Anthropic (Jul 24, 2026).