Anthropic Launches Claude Opus 5, Bringing Near-Frontier AI Performance at Half the Cost

Anthropic has unveiled Claude Opus 5, its latest flagship AI model, at a significantly lower cost than its most advanced model, Claude Fable 5. Anthropic has priced the model at $5 per million input tokens and $25 per million output tokens, matching the pricing of Opus 4.8.

The company said Opus 5 offers state-of-the-art performance on coding and knowledge work while costing half as much as Fable 5, making it its new default model for Claude Max subscribers and the strongest model available on Claude Pro.

According to Anthropic, Opus 5 sets a new benchmark on software engineering tasks. The model outperformed rival AI systems on Frontier-Bench v0.1, more than doubling the performance of its predecessor, Opus 4.8, at a lower cost per task. On CursorBench 3.2, Opus 5 scored within 0.5% of Fable 5's peak performance while requiring only half the cost.

The company also highlighted improvements in reasoning and enterprise automation. On ARC-AGI 3, which measures a model's ability to solve novel problems, Opus 5 scored three times higher than the next-best model.

It also led AutomationBench, a benchmark that evaluates AI's ability to complete end-to-end business workflows, and outperformed competitors on OSWorld 2.0, a computer-use benchmark.

Beyond coding, Anthropic said Opus 5 delivers meaningful gains for scientific research. The model demonstrated improved performance across life sciences, including structural biology, bioinformatics, and organic chemistry. It also showed stronger capabilities in generating complex visual outputs and handling multimodal tasks.

Anthropic said Opus 5 is better at verifying its own work and iterating until it reaches a correct solution. During internal testing, the model successfully created its own computer vision pipeline to reconstruct a 3D machine part from raw image pixels after being intentionally denied direct access to the original drawing. In another test, it identified and fixed the root cause of a bug in a widely used open-source package manager that other AI models failed to resolve.

On safety, Anthropic said Opus 5 is its most aligned model to date, exhibiting lower rates of deceptive behavior and greater resistance to misuse than previous Claude models.

While the model has improved in identifying software vulnerabilities, the company emphasised that it remains behind Mythos 5 in offensive cybersecurity tasks such as exploit generation.

To reduce misuse, Anthropic has implemented stricter safeguards that block binary-based vulnerability scanning, penetration testing, and exploit generation while continuing to allow legitimate cybersecurity research.

Claude Opus 5 is available starting today across all Claude platforms.

A faster version, capable of running approximately 2.5 times faster, is also available for developers through the Claude API.