Claude Opus 5
Claude Opus 5 is Anthropic’s newest frontier AI model aimed at coding, research, business work and other complex tasks. The company says it represents a major performance jump from Opus 4.8 while keeping the same API pricing.
Anthropic also claims Opus 5 approaches Claude Fable 5 on several coding and computer-use benchmarks while delivering a much lower AI cost per task. The model is available across Claude’s apps and API, and it’s now the default for Claude Max subscribers and the most capable option included with Claude Pro.
Benchmark performance and cost
On CursorBench 3.2, which evaluates coding agents on real software development tasks, Opus 5 finished within 0.5% of Fable 5’s best score at maximum effort. Anthropic says it achieved that result at roughly half the cost per task.
- CursorBench: Opus 5 scored within 0.5% of Fable 5 at max effort, at about half the cost per task.
- OSWorld 2.0: Opus 5 outperformed Fable 5’s best result while running at just over a third of the cost. This benchmark gauges how well AI agents operate computers and complete multi-app, multi-file tasks.
For Pro and standard Team subscribers, Opus 5 could help fill the gap left when Fable 5 moved behind pay-as-you-go credits on July 20.
Improvements over Opus 4.8
Anthropic reports Opus 5 more than doubled Opus 4.8’s score on Frontier-Bench, which tests AI agents on difficult software engineering challenges, and completed those tasks at a lower average cost. The company highlights improvements in self-checking, bug diagnosis, and persistence through tough jobs rather than stopping after a superficial fix. In one example, Opus 5 identified an edge case missed by an existing community patch.
These results come from Anthropic’s internal testing, so real-world performance may vary across everyday workloads. Pricing remains unchanged from Opus 4.8: $5 per million input tokens and $25 per million output tokens.
