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

Anthropic on Monday, November 09, released two new AI models, Claude Fable 5.2 and Claude Mythos 5.2. The Dario Amodei-led company is releasing its frontier models for the first time since its call for a slowdown of AI development. The new models also come close on the heels of the company’s preparations for a historic IPO, which is likely to take place sooner.

Anthropic claims that both Fable 5.2 and Mythos 5.2 are its most advanced offerings. The models are said to generate output up to 30 per cent faster than their predecessors and also cost about 20 per cent less than Fable and Mythos 5.1. The new models are rolling out with a price tag of $8 per million input tokens and $40 per million output tokens.

Further, the company said that Claude Fable 5.2 and Claude Mythos 5.2 show improvements in coding, reasoning, and vision. During testing ahead of release, the models helped Anthropic engineers to accomplish difficult software migrations, design three working electronic devices from scratch, solved two open problems in physics of black holes, made considerable progress on robotics evaluations, and formalised the proof of Fermat’s Last Theorem in 30 hours, etc.

Anthropic said that Fable 5.2 and Mythos 5.2 are based on the same underlying AI model; however, they differ in terms of their safety restrictions and availability. Fable 5.2 features safeguards that limit potentially dangerous cybersecurity and biological research applications, making it available to a wider audience. On the other hand, Mythos 5.2 has fewer restrictions and is meant for vetted organisations working on cybersecurity and biomedical research. According to the frontier AI lab, the models are being designed for ambitious, open-ended technical projects that need sustained reasoning and the capability to manage multiple AI agents.

When it comes to performance, the company said that Fable 5.2 outperformed its predecessor, Fable 5.1, and scored higher than several rival models on most of its published evaluations. On FrontierCode 2.0, a benchmark assessing agentic software development, Fable 5.2 scored 49.1 per cent, compared with 45.5 per cent for Claude Opus 5.5 and 42.8 per cent for OpenAI’s GPT-6 Astra. It also scored 69.9 per cent on FrontierSWE v2, compared with 62.3 per cent for Opus 5.5.

The model recorded 74.6 per cent on Terminal-Bench-Science, which evaluates AI agents on scientific research tasks. Its Mythos 5.2 counterpart scored 75.3 per cent. Fable 5.2 also achieved 89.3 per cent on OSWorld 2.1, a computer-use benchmark, and 64.9 per cent on RoboDojo, which tests robotic control. These figures come from Anthropic’s own evaluations, and it is to be noted that benchmark results do not necessarily translate into similar performance across everyday applications.

Both models are aimed at applications that go beyond conventional AI chatbot interactions. In software development, Fable 5.2 can handle multiple AI agents to write, test and review code. Anthropic showed this by using more than 1,000 agents to build a working operating system over eight days. In scientific research, the model helped formalise the proof of Fermat’s Last Theorem in 30 hours using 1,000 agents. It also helped researchers solve two open problems in theoretical physics involving black holes.

Story continues below this ad

According to the company, Mythos 5.2 is particularly suited for cybersecurity research, identifying software vulnerabilities and helping security teams fix them. In biology, it successfully modelled 28 of 34 protein structures associated with adeno-associated viruses, which are used in gene therapy research.

Claude Fable 5.2 is available now across Anthropic’s platforms, including Claude, Claude Code and its API. Developers can also access it through Amazon Web Services, Google Cloud and Microsoft Azure. Mythos 5.2, meanwhile, is not generally available yet.