Anthropic released Claude Opus 5 on 24 July 2026, and the headline number is the price tag, not the benchmark scores: $5 per million input tokens and $25 per million output tokens, identical to the outgoing Opus 4.8, while the model itself closes most of the gap to Anthropic’s most capable system, Claude Fable 5. On CursorBench 3.2 at maximum effort, Opus 5 lands within 0.5% of Fable 5’s peak agentic-coding score at roughly half the cost per task. On OSWorld 2.0 — a computer-use benchmark that tests whether a model can actually operate software rather than just describe it — Opus 5 beats Fable 5’s best result at just over a third of the cost. It also scored 26.0% on AutomationBench and 30.16% on ARC-AGI-3, carries a 1M-token context window, and is live the same day across Claude.ai, the Claude API, Claude Code, and Claude Cowork. It’s now the default model on Claude Max and the strongest model available on Claude Pro.
The other detail worth noting is the “effort dial” that shipped alongside it — a control that lets a team trade latency and cost against task difficulty on a per-request basis, rather than picking one model tier and living with its price for every job, trivial or not.
So what
The practical read for anyone commissioning AI-assisted software work isn’t “which model is smartest” — it’s that the cost of running frontier-tier agentic coding capability in production just dropped without a corresponding drop in quality. Six months ago, the choice was between a cheaper model that made more mistakes on multi-step agentic tasks and an expensive one that didn’t. That trade-off is compressing. For teams weighing whether AI-assisted development is viable for anything beyond prototyping — sustained agentic workflows, computer-use automation, long-running coding tasks — this is exactly the kind of pricing shift that changes the answer. If you’re evaluating how much of your build should lean on AI tooling like this versus senior engineering judgment, that’s the trade-off we help clients make on AI-assisted development projects, or get in touch to talk through where the line should sit for yours.