Anthropic has officially introduced Claude Opus 5, a new artificial intelligence model designed for programming, professional analysis, and complex business tasks. The main goal of the model is to deliver performance close to the company’s flagship model, Claude Fable 5, while reducing operating costs by nearly half. Instead of focusing only on higher benchmark scores, Anthropic aims to provide a more cost-effective AI solution for businesses and developers that process millions of tokens every day.
Claude Opus 5 became available on July 24, 2026, through the Claude app and the Claude API. The model costs $5 per million input tokens and $25 per million output tokens, making it significantly cheaper than Claude Fable 5. It is now the default model for Claude Max users and the most powerful option included in the Claude Pro subscription. The new model is designed for demanding tasks such as software development, document analysis, long-running AI agents, and enterprise automation. Anthropic says Claude Opus 5 is more careful before responding because it reviews its own reasoning, checks for mistakes, and corrects errors when necessary. This improves consistency during long and complicated workflows rather than simply producing fast answers.
Developers can access Claude Opus 5 through the Claude API, Amazon Bedrock, Google Cloud, and Microsoft Foundry. The model supports a context window of up to one million tokens, can generate as many as 128,000 output tokens, and allows users to control the level of reasoning effort depending on the complexity of a task. This flexibility helps reduce computing costs for simple requests while providing deeper analysis for scientific research, financial work, or software architecture.
Anthropic claims Claude Opus 5 performs almost as well as Claude Fable 5 in programming benchmarks while costing nearly half as much. According to the company’s published tests, it scored only slightly below Fable 5 in CursorBench but achieved the results at a much lower cost. In OSWorld benchmarks, Anthropic says Opus 5 even outperformed Fable 5 in some tasks while using significantly fewer resources. Although these results come from Anthropic’s own evaluations, they indicate the company’s focus on improving efficiency instead of only increasing raw performance.
Software engineering is one of the model’s biggest strengths. Claude Opus 5 can write code, identify the root cause of software bugs, generate verification tests, and automatically revise its solutions when errors are detected. Anthropic also demonstrated the model rebuilding a mechanical component in FreeCAD by creating its own computer vision process to interpret an image that it could not directly access. In another demonstration, it successfully identified and fixed a complex issue in an open-source package manager beyond previously suggested community solutions.
The model also brings improvements to financial analysis, legal document review, scientific research, business automation, and presentation creation. Early enterprise users reported better numerical accuracy, fewer unnecessary processing steps, and more consistent results across long projects. These improvements make Claude Opus 5 particularly attractive for organizations that require reliable AI assistance for professional work.
Anthropic has also refined the model’s security system. Claude Opus 5 is better at identifying software vulnerabilities while including stronger safeguards to prevent misuse for offensive cyberattacks. The company says its safety filters interfere less with legitimate defensive security work while continuing to block exploit generation and other harmful activities. When a request exceeds the model’s safety limits, Anthropic can automatically redirect it to another compatible model instead of ending the conversation.
Another important improvement is the introduction of new developer features. Developers can switch tools during an ongoing conversation without losing cached data and configure automatic fallback models if a request becomes blocked. These workflow improvements are especially valuable for enterprise applications where uninterrupted AI operations are critical.
Overall, Claude Opus 5 is not intended to completely replace Claude Fable 5. Instead, it offers a practical balance between performance and cost by delivering near flagship-level capabilities at a significantly lower price. For businesses processing large volumes of AI requests, the combination of lower costs, improved reasoning, stronger coding abilities, and better workflow management makes Claude Opus 5 one of Anthropic’s most important AI releases to date.
