Anthropic Launches Claude Fable 5.2 and Mythos 5.2 Models

What Happened

Anthropic has announced two new models, Claude Fable 5.2 and Claude Mythos 5.2, which it says are designed to take on “daunting” technical challenges [1]. According to Anthropic, the models show improvements in coding, reasoning and vision, and they generate outputs faster than the 5.1 versions of Fable and Mythos [1]. TechCrunch reports that Anthropic also says the models communicate more clearly with users [2].

Claude Fable 5.2 is launching today. Anthropic says it is good at challenging software engineering problems and can orchestrate agent teams and tackle long-running projects [1]. Claude Mythos 5.2 follows the pattern of earlier Mythos releases: only vetted organizations can use it to start [1]. The new 5.2 models follow the release of Claude Fable 5.1 and Claude Mythos 5.1 in September [1].

During prerelease testing, Anthropic says the models helped its engineers complete software migrations that would otherwise have been infeasible, and designed three working electronic devices from scratch [1]. Anthropic also says the models solved two open problems in the physics of black holes, made substantial progress on robotics evaluations, and formalized the proof of Fermat’s Last Theorem in 30 hours [1]. TechCrunch reports that Fable 5.2 had been available to select parties in prerelease for some time, and Anthropic says it solved two open problems in theoretical physics [2].

On pricing, TechCrunch reports that Anthropic is lowering the cost of Fable from $10 to $8 per million input tokens. As a result, TechCrunch says Fable 5.2 is now cheaper than its primary competitor, GPT-6 Astra [2]. Anthropic says the new Fable returns responses 30% faster across effort levels, and faster inference serving is available at double the price [2].

On safety and testing, TechCrunch reports that a third-party report from the assessment group METR suggested Anthropic might be approaching recursive self-improvement, meaning AI models carrying on model training independently. The METR report found internal models were speeding up Anthropic’s AI progress by roughly 2x, although many uncertainties remain [2]. Anthropic says both 5.2 models voice suspicion that they are being evaluated more often than its recent models, which challenges its ability to assess how they will act in real-world settings [2].

The Mythos 5.2 system card says the model shows stronger covert capabilities than prior Claude models on several evaluations, particularly when its reasoning is hidden from the monitor. The same card says low levels of absolute performance on the new covert evaluations support claims that rely on detecting harmful actions in high-stakes situations [2]. Anthropic maintains that the new Mythos does not have the ability to act covertly against its user [2].

On the same day, AI startup Physical Superintelligence announced a new proof of the Aaronson-Ambainis conjecture, a step in setting theoretical limits for quantum computing. The lab said it achieved the result by running “a non-public Anthropic model” within its own system. Anthropic did not confirm or deny whether Fable 5.2 was that model [2].

Who Is Affected

• Vetted organizations: Mythos 5.2 can only be used by vetted organizations to start [1].
• Enterprise clients: TechCrunch reports that Fable 5.2 is compatible with zero data retention policies through the Enterprise Frontier Safeguards system, which it calls a long-standing sticking point for many enterprise clients [2].
• Customers paying per token: Fable’s input-token price is being cut from $10 to $8 per million tokens [2].