Anthropic has announced Claude Opus 5, its latest generative AI model, introducing significant improvements in coding, reasoning and enterprise automation while maintaining the same API pricing as its predecessor.
According to the company, Claude Opus 5 delivers performance close to its flagship Claude Fable 5 model across several benchmarks, offering developers and businesses a more cost-effective option without compromising on capability.
Claude Opus 5 delivers stronger performance across key benchmarks
Anthropic says Claude Opus 5 achieved leading results in several industry benchmarks designed to evaluate coding, reasoning and computer interaction.
Some of the reported highlights include:
- More than double the Frontier-Bench v0.1 score achieved by Claude Opus 4.8 while operating at a lower cost.
- A CursorBench 3.2 score within 0.5% of Claude Fable 5’s maximum-effort performance while costing roughly half as much.
- Around three times the score of the next-best model in the ARC-AGI 3 benchmark, which measures the ability to solve unfamiliar reasoning problems.
- Approximately 1.5 times the pass rate of the next-best model at the same cost in Zapier AutomationBench, which evaluates business task completion.
- Leading performance in OSWorld 2.0, a benchmark that measures a model’s ability to operate a computer, outperforming other competing models at every cost level.
The company also reported improvements in scientific tasks, including a 10.2-point increase in organic chemistry spectral analysis and a 7.7-point gain in protein sequence function prediction compared to Claude Opus 4.8.
Improved real-world AI agent capabilities
Beyond benchmark scores, Anthropic highlighted several real-world examples demonstrating Claude Opus 5’s capabilities.
In one demonstration, the model reportedly created its own computer vision pipeline to analyse engineering drawings and reconstruct a 3D FreeCAD model without requiring direct visual inspection of the original design.
Anthropic also shared feedback from an engineer at a trading company who successfully used Claude Opus 5 to build a new market data feed for an exchange in a single session. According to the company, previous AI models struggled to complete the same task, even with detailed planning.
Stronger safety and alignment
Anthropic says Claude Opus 5 is its most aligned AI model to date based on the company’s internal behavioural evaluations.
The company reported stronger adherence to Claude’s Constitution, improved resistance to deceptive behaviour and lower susceptibility to misuse compared to Claude Opus 4.8, Sonnet 5 and Claude Fable 5.
While Anthropic says Opus 5 performs well in cybersecurity-related evaluations, it notes that the model still trails Mythos 5 in certain advanced cybersecurity and biology tasks, particularly exploit generation.
To reduce potential misuse, Anthropic has restricted capabilities related to binary vulnerability scanning, penetration testing and exploit generation. However, approved researchers participating in the company’s Cyber Verification Program (CVP) can access a less restrictive version of the model.
New developer features
Alongside Claude Opus 5, Anthropic introduced two beta features for developers:
- Tool changes during conversations: Developers can now change the tools available to Claude during an ongoing conversation without invalidating the prompt cache, making long-running AI workflows more efficient.
- Automatic fallback: Requests blocked by safety filters can now be automatically redirected to another compatible model instead of being rejected outright, improving application reliability.
Pricing and availability
Claude Opus 5 is available immediately through the Claude API under the model name claude-opus-5.
Anthropic has kept pricing unchanged from Claude Opus 4.8:
- $5 per million input tokens
- $25 per million output tokens
The model is available on Anthropic’s paid Claude plans, including Pro and Max. Free-tier users continue to have access only to Claude’s standard features and do not receive access to Opus models.
What is Claude Opus 5?
Claude Opus 5 is Anthropic’s latest high-performance AI model designed for complex coding, reasoning, automation and scientific workloads. It targets developers, researchers and businesses that require advanced AI capabilities while offering improved cost efficiency compared to the company’s flagship models.
With stronger benchmark performance, enhanced real-world agent capabilities, improved safety measures and unchanged pricing, Claude Opus 5 represents Anthropic’s latest effort to make advanced AI more capable and accessible for enterprise users.
