Anthropic Releases Claude Fable 5.2 and Mythos 5.2: Models That Write Operating Systems and Design Electronics on Their Own

On November 9, Anthropic released two new models, Claude Fable 5.2 and Mythos 5.2. The company positions them for complex engineering, research, and scientific tasks. According to Anthropic’s internal tests, both models outperform Claude Opus 5.5 in most evaluated areas.

In one experiment, Fable 5.2 developed a functioning operating system from scratch. The model coordinated more than 1,000 AI agents, who spent eight days writing, testing, and verifying the code. The result was an OS capable of running on a laptop, complete with a file system, drivers, a graphical interface, and an installer. On the FrontierCode 2.0 benchmark, the model scored 49.1%, compared to 45.5% for Opus 5.5 and 42.8% for GPT-6 Astra.

A team of 1,000 Fable 5.2 agents formalized the proof of Fermat’s Last Theorem in 30 hours—that is, they translated it into a format where a computer can verify each step. A previous research version of Claude took 11 days to accomplish this. Anthropic also reports that the model solved two previously unsolved problems in theoretical physics related to black holes and quantum gravity. Based on brief text instructions, Fable 5.2 designed three electronic devices: a home voice assistant, a USB device for conference calls, and a portable toy. All three worked after assembly without any modifications to the electrical circuits.

The second model, Mythos 5.2, is designed to detect and eliminate software vulnerabilities. In the CVE Rediscovery test, it detected 89.6% of known vulnerabilities, compared to 78.3% for Mythos 5.1. Access to Mythos is restricted to verified organizations working in cybersecurity and biomedical research.

The new models generate responses up to 30% faster than previous versions. The API costs $8 per 1 million input tokens and $40 per 1 million output tokens; there is also an accelerated mode that generates responses up to 2.5 times faster at double the price. Fable 5.2 is available to Claude users, while Mythos 5.2 is provided through special verification programs. Most of the results cited were obtained in Anthropic’s experiments and still require independent verification.