WORLD - Artificial intelligence company Anthropic has released Claude Fable 5.2 and Claude Mythos 5.2, describing them as its most advanced models yet, with improved abilities to write computer programs, solve scientific problems and carry out complex tasks.
The two models are designed to handle tasks that require several steps and coordinate multiple AI systems, moving beyond simply answering questions or generating text.
Coding and Scientific Achievements
According to Anthropic, Fable 5.2 scored 49.1% on the FrontierCode 2.0 test, which measures how well AI models handle complex programming tasks. The company said the result surpassed scores achieved by OpenAI's GPT-6 Astra and Google's Gemini 4 Carbon.
In internal experiments, Fable 5.2 coordinated more than 1,000 AI agents to build a functioning operating system in eight days. An operating system is the software that manages a computer's basic functions.
Anthropic also reported that the model completed a formal mathematical proof of Fermat's Last Theorem in 30 hours, helped solve two previously unanswered problems in black hole physics and designed three working electronic devices.
The company further reported improvements in the model's ability to perform tasks related to robotics.
Anthropic's reported results are based on its own experiments and have not all been independently verified.
Speed, Pricing and Access
Both models can produce answers up to 30% faster than their predecessors, according to Anthropic.
Fable 5.2 is available to users, with API prices starting at $8 per million input tokens. An API allows developers to connect AI models to their own applications and services.
Mythos 5.2, meanwhile, is restricted to verified organizations working in sensitive research fields.
The releases reflect growing efforts by AI companies to develop systems capable of handling more complex technical and scientific work, rather than focusing primarily on everyday conversations and content generation.
