Anthropic has released Claude Fable 5.2 and Claude Mythos 5.2, bringing advanced capabilities in coding, reasoning, and cybersecurity engineering.
Developers can explore improvements across benchmarks such as CyScenarioBench and CVE Rediscovery, alongside new Claude Security features including continuous code scanning and automated fix tracking.
According to the company, during the testing period before their release, the models helped complete code migrations, designed three working electronic devices from scratch, solved two black hole physics problems, made progress in robotics, and came up with a proof of Fermat's Last Theorem in 30 hours.
Anthropic said the new models performed at or above previous versions, such as the Claude Opus 5.5, Sonnet 5.5 and Mythos 5.1, in areas such as alignment, abuse resistance, fidelity and accuracy.
Digital security testing and performance
In a test of simulated attacks, Fable 5.2 completed 76,3% of the challenges, beating Fable 5.1’s 61,7% and OpenAI’s GPT-6 Cyber’s 61%. In the same test, Google’s Gemini 4 Argon solved 12 challenges from start to finish, an improvement over its previous versions. OpenAI’s model also completed 11 long-running scenarios with an average success rate of 61% and solved all of the smaller challenges.
↳
READ ALSO
HACKS.GR / RELATED
AI Unlocked
Otter.ai gains support for 31 languages and automatically understands who is speaking
↗
Additionally, in a test of discovering known security issues based on published reports from October 2026, Fable 5.2 scored 89,6%.
Anthropic described Claude Mythos 5.2 as its most capable cybersecurity model yet, capable of detecting, reproducing, and fixing more problems.
Availability and new protection tools
Claude Fable 5.2 is available on major online service platforms, while its security features are offered through a special verification program.
At the same time, the company updated its security service, adding continuous code scanning and automatic patch tracking. Each finding is accompanied by a severity rating and an explanation of how someone could have reached the problematic point in the code.
>Stay connected with Hacks.gr
Set us as a preferred source on Google and find us on social media.
Preferred source
↗
end of article
Top comments (0)