Anthropic unveils Claude Fable 5.2 and Claude Mythos 5.2 ahead of Nasdaq listing

In scientific research evaluation using the Terminal-Bench-Science 0.1 benchmark, Claude Fable 5.2 achieved 74.6 percent and Claude Mythos 5.2 achieved 75.3 percent, compared with 52.6 percent for the previous Claude Fable 5.1. Claude Fable 5.2 solved two previously unsolved theoretical physics problems related to black holes and quantum gravity theory. The model performed calculations demonstrating that phase transitions in black hole formation occur sharply even in very small systems, and identified limitations in mathematical techniques used to validate quantum gravity theory candidates. Anthropic presented both research results as preprints.

In the life sciences field, Claude Mythos 5.2 developed a new computational method for predicting capsid structures of adeno-associated viruses used in gene therapy. The approach combined existing open-source protein structure prediction models and leveraged symmetry in protein complexes. In Anthropic's testing, the model accurately modeled 28 out of 34 target capsid structures.

For robotics control, Claude Fable 5.2 scored 64.9 percent on the RoboDojo Sim benchmark, which evaluates a two-armed robot's ability to manipulate objects based on camera images. This represents a 37.3 percentage point increase from Claude Fable 5.1's 27.6 percent score. The model demonstrated capabilities in solving puzzle cubes, opening bottle caps, and drawing. Anthropic also reported that Claude Fable 5.2 designed three functioning electronic devices including a voice assistant hub, a USB speakerphone, and a portable virtual pet device, handling component selection, circuit board design, 3D printing model creation, and software development. All three devices operated without electrical design modifications after physical manufacturing.

In mathematics, approximately 1,000 Claude Fable 5.2 agents collaborated to completely formalize the proof of Fermat's Last Theorem in 30 hours. This formalization process converts existing proofs into computer-verifiable mathematical language. The effort used approximately one-third the action tokens of a previous research model and produced a formalized proof seven times shorter than earlier versions.

In software development, Claude Fable 5.2 coordinated over 1,000 sub-agents to develop an operating system prototype over eight days, implementing bootable functionality, storage, keyboard, network operations, file storage, window management, login screens, and installation programs. On the FrontierCode 2.0 agentic coding benchmark, Claude Fable 5.2 scored 49.1 percent compared with 32.4 percent for the previous model. On the OSWorld 2.1 computer use benchmark, it achieved 89.3 percent versus 83.5 percent previously.

For cybersecurity, Claude Mythos 5.2 scored 89.6 percent on the CVE Rediscovery evaluation, which measures whether an artificial intelligence model can independently identify known software vulnerabilities in code from before patches were released. This represents an 11.3 percentage point increase from Claude Mythos 5.1's 78.3 percent. Anthropic evaluated 1,829 vulnerabilities in October 2026. In testing by a security firm, approximately one-sixth of vulnerabilities found by Claude Mythos 5.2 were not discovered by other frontier artificial intelligence models.

Output generation speed improved by up to 30 percent compared with previous versions. The base application programming interface pricing is set at 8 dollars per million input tokens and 40 dollars per million output tokens. An optional fast mode increases token generation speed by up to 2.5 times, priced at 16 dollars per million input tokens and 80 dollars per million output tokens.

Anthropologic submitted its securities registration statement for the initial public offering to the United States Securities and Exchange Commission on October 26. The number of shares offered and price range have not yet been determined. The model release timing coincides with the company's Nasdaq listing preparations, highlighting technological competitiveness as investors evaluate artificial intelligence companies based on revenue growth, computing investments, and business expansion potential.