Anthropic unveils Claude Fable 5.2, Mythos 5.2 ahead of potential IPO

Anthropic, a company led by Dario Amodei, announced the release of two new AI models, Claude Fable 5.2 and Claude Mythos 5.2, on November 9. These models represent Anthropic's most advanced offerings to date, released just before the company anticipates a historic IPO.

Both Fable 5.2 and Mythos 5.2 demonstrate significant improvements over their predecessors, generating output up to 30% faster and costing about 20% less than Fable and Mythos 5.1. They are priced at $8 per million input tokens and $40 per million output tokens.

The new models exhibit enhanced capabilities in coding, reasoning, and vision. During testing, they assisted Anthropic engineers in tackling complex software migrations, designing electronic devices from scratch, and progressing on physics and robotics evaluations. Remarkably, they even helped formalize the proof of Fermat’s Last Theorem in 30 hours.

Fable 5.2 includes safety restrictions that limit its use to cybersecurity and biological research applications, making it more accessible to a broader audience. In contrast, Mythos 5.2 has fewer restrictions and is reserved for vetted organizations focusing on cybersecurity and biomedical research.

In terms of performance, Fable 5.2 outperformed its predecessor, scoring higher than several rival models on most published evaluations. On FrontierCode 2.0, it scored 49.1%, compared to 45.5% for Claude Opus 5.5 and 42.8% for OpenAI's GPT-6 Astra. It also excelled in FrontierSWE v2 and Terminal-Bench-Science benchmarks. Similarly, Mythos 5.2 scored 75.3% on OSWorld 2.1 and 64.9% on RoboDojo benchmarks, showcasing its prowess in scientific research and robotic control.

These AI models are designed for ambitious, open-ended technical projects requiring sustained reasoning and the ability to manage multiple AI agents. While both models are capable of handling multiple AI agents to write, test, and review code, they differ in their application areas. Fable 5.2 is well-suited for software development, scientific research, and cybersecurity research, while Mythos 5.2 is particularly advantageous for cybersecurity and biomedical research.