Anthropic, a leading artificial intelligence company, has introduced two new models, Claude Fable 5.1 and Mythos 5.1. These are essentially the same model but equipped with different security features tailored to specific use cases. The updates significantly reduce the cost of using the models by 75 percent, with a price of $0.25 per million tokens. This cost-cutting could reduce typical workloads by 25 percent and more complex "agentive" tasks by up to 45 percent. While Fable 5.1 is available to the general public, Mythos 5.1 is restricted to select users through specialized access programs in cybersecurity and life sciences. The life sciences program was developed in collaboration with the U.S. government. One notable feature of the new models is the inclusion of a watermark compliant with the European Union's AI Act. This watermark helps identify AI-generated content and is currently in a private preview for eligible organizations, including regulatory bodies, law enforcement, and academic institutions. Anthropic signed the EU Code of Good Practices on AI transparency in July, joining over 190 other organizations. The watermark is invisible to the naked eye and does not contain user-specific information. While Anthropic had previously planned to offer a free detection API, the current version is more restricted. Anthropic has not specified a timeline for expanding access to its models but has indicated it intends to do so over time. The second European gap refers to the limited access to Mythos 5.1, which remains available only to a select group of American organizations, two years after the launch of Claude in the EU. The company aims to integrate more international partners, but currently, European researchers in cybersecurity and life sciences are not included in the program. Anthropic is also preparing for a major milestone: its initial public offering (IPO). In August 2026, the company planned to file the necessary documents for its IPO by the end of the month. This IPO could potentially match or exceed the $75 billion raised by SpaceX in June, a record for the technology sector. The filing would provide the first detailed look at the company's financials and could give Anthropic an advantage in the stock market over competitors like OpenAI. Claude Fable 5.1 has set a new benchmark in performance across various fields, including programming and long-haul problem-solving. It achieves significantly better results than its predecessor, Fable 5, and can perform at a similar or better level at a lower cost when set to a "Low" or "Medium" effort level. Fable 5.1 is also more efficient at identifying root causes of software problems, as demonstrated by its successful identification of a rare system crash that had puzzled engineers for years. In scientific research, both Fable 5.1 and Mythos 5.1 have shown remarkable capabilities. Mythos 5.1, for example, was able to design high-affinity binding molecules with a success rate of nearly 50 percent, significantly higher than typical rates in protein design. Additionally, Fable 5.1 created a high-resolution topographic map of Venus using historical data, improving the scale and accuracy of the map. In computational biology, Mythos 5.1 optimized deep learning models, reducing the time and cost of research tasks. Safety and reliability remain central to Anthropic's development. Despite the increased capabilities of AI models, the company emphasizes that safety, reliability, and alignment must keep pace with these advancements. Evaluations show that Mythos 5.1 is not yet at a high-risk level for chemical or biological weapons production, and its cybersecurity features have been rigorously tested. The company continues to refine its models to ensure they are both powerful and secure, aiming to support scientific progress while minimizing risks.