Anthropic launches Claude Fable and Mythos 5.2 models

New models for technical tasks

Anthropic has officially introduced two new artificial intelligence models: Claude Fable 5.2 and Claude Mythos 5.2. These models are built to handle sophisticated technical challenges that require high levels of logic and precision. The company notes that the new versions provide significant improvements in coding, reasoning, and visual understanding.

The developers state that these models are designed to tackle the most daunting technical challenges. They aim to provide a more capable partner for users working on ambitious projects across various professional fields. By improving these core capabilities, Anthropic is positioning the new models for heavy-duty industrial and scientific use.

According to Digit, the models are expected to handle complex tasks more efficiently than their predecessors. This includes specific applications in software development, cybersecurity-related work, and scientific research. The goal is to provide a tool that can navigate the nuances of professional workflows while maintaining high accuracy.

Benchmarks in physics and software

Anthropic shared several internal test results to demonstrate the capabilities of the new models. One notable achievement involved the models designing three working electronic devices from scratch. Additionally, the models solved two previously unsolved problems in the field of theoretical physics.

The company also highlighted the models' ability to perform complex computer-based formalization. Specifically, the models completed the formalization of Fermat’s Last Theorem in just 30 hours. These results suggest a high level of proficiency in mathematical logic and structured problem-solving.

Another large-scale test involved 1,000 Fable 5.2 agents working together to build a functional operating system. This project took approximately eight days to complete. The resulting system included a core, hardware drivers, file storage, windows, a login screen, and an installer.

Speed and efficiency gains

The new models are not just more capable; they are also designed to be faster and more economical. Anthropic claims that the models can generate responses up to 30 per cent faster than the previous Claude Fable 5.1 and Mythos 5.1 versions. This speed is a key factor for users who need real-time interactions or high-volume processing.

Efficiency is also measured by the number of tokens—the basic units of text processed by the AI—required to complete a task. The new models are designed to use fewer tokens for certain complex tasks, which helps in managing costs and processing limits. They are also built to communicate more clearly with human users.

Anthropic has adjusted the pricing for both models to be more competitive compared to previous versions. The standard pricing starts at $8 per million input tokens and $40 per million output tokens. For users who prioritize speed over cost, a fast mode is available at $16 per million input tokens and $80 per million output tokens.

Access and availability

While both models represent a step forward for the Claude family, their availability differs significantly. Claude Fable 5.2 is currently available to all Claude users. This allows a wide range of individuals to experiment with the new reasoning and coding capabilities immediately.

In contrast, access to Claude Mythos 5.2 is currently restricted. Anthropic is only offering this specific model to verified organizations that work in cybersecurity and biomedical research. This suggests that Mythos 5.2 may contain specialized capabilities or safety guardrails tailored for high-stakes environments.

By separating the availability of these two models, Anthropic can manage the deployment of its most specialized tools. Organizations in the biomedical and security sectors will have access to a model tuned for their specific needs. Meanwhile, the general public can utilize the Fable 5.2 model for a broader range of creative and technical applications.

Future of AI collaboration

The introduction of Fable and Mythos 5.2 marks a move toward AI as a collaborator on ambitious projects. By showing improvements in vision and reasoning, Anthropic is expanding what these models can do beyond simple chat interfaces. They are becoming tools that can contribute to the actual construction of systems and the solving of theoretical problems.

The ability of 1,000 agents to work together on an operating system is a significant milestone for multi-agent systems. This type of collaborative AI can potentially automate large portions of software engineering. It allows for a modular approach where different agents handle different components of a larger project.

Moving forward, the focus will likely remain on refining these technical capabilities for professional use cases. As the models become faster and more efficient, they will become more viable for real-time industrial applications. The restriction on Mythos 5.2 indicates that specialized, high-security applications remain a primary frontier for the company.