Anthropic has officially launched Claude Opus 5.5, a new artificial intelligence model designed to deliver powerful capabilities at a reduced cost. This latest iteration offers similar advanced features and safety enhancements as the Claude Fable 5.1 model but comes with a significant 40% cost reduction compared to the earlier Opus 5, which debuted in July.
The company highlights that Opus 5.5 excels particularly in agentic tasks, complex coding, general computer use, and various forms of knowledge work, making it a versatile tool for developers and businesses.
Enhanced Performance and Benchmarks
Early testers have reported notable improvements across intricate workflows. One remarkable example cited by Anthropic involved an early tester completing a 680,000-line code migration in under a day, a task that would typically occupy an engineering team for several weeks. The model also demonstrates a quicker ability to identify and rectify software inefficiencies.
In benchmark tests, Claude Opus 5.5 has shown superior performance against several competing AI models. In GDPval-AA v2 testing, which assesses AI performance in real-world work scenarios, Opus 5.5 achieved a score of 1,846, surpassing Claude Fable 5.1's 1,735. Furthermore, it outperformed GPT-6 Astra on Terminal-Bench 4.0, scoring 66.4% compared to Astra's 57.9%, indicating an 8.5% higher efficiency. However, GPT-6 Astra maintained an edge in Terminal-Bench-Science 0.1.
Commitment to Safety and Future Developments
Anthropic emphasized its rigorous safety protocols for Claude Opus 5.5, including extensive internal testing and evaluations by external organizations like METR and Frontier Design. The model is engineered to be significantly less prone to taking irreversible actions or operating outside defined boundaries. It also exhibits enhanced resistance to prompt injection compared to its predecessor, Opus 5.
The company is actively developing more robust safety and monitoring systems for its upcoming AI models. These efforts include preparing stronger protections against potential misuse or harmful behaviors, alongside new methods for safe model training and comprehensive risk evaluation prior to deployment. Anthropic plans to share more details on these initiatives soon, underscoring the importance of public policy in ensuring the safety of increasingly capable AI systems.
Availability and Upcoming Models
Claude Opus 5.5 is now widely accessible across major cloud platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure. Developers can also utilize the model directly through the Claude Platform. Looking ahead, Anthropic has announced plans to launch Claude Sonnet 5.5 and Claude Haiku 5.5 models in the near future, further expanding its AI offerings.