Anthropic Introduces New Fable 5.2 and Mythos 5.2 AI Models
Overkill for the average user, but the models are smarter than ever for professionals.
On Monday, Anthropic unveiled its latest AI models, Fable 5.2 and Mythos 5.2, promising significant advancements over their predecessors. Though the average user may not immediately notice the benefits, Anthropic claims these new models are ideal for complex, open-ended tasks.
The Claude models demonstrate marked improvements in coding, reasoning, vision, cybersecurity, and biology. Particularly, the Mythos 5.2 model boasts robust safeguards against misuse, instilling Anthropic with confidence to release it publicly. Fable 5.2 and Mythos 5.2 generate output up to 30% faster than their predecessors and now come at more affordable prices, a trend observed in recent AI model releases.
Thorough benchmarks prove the new models' superiority over their predecessors. Fable 5.2 outperforms Fable 5.1 by 49.1% to 32.4% in an agentic coding benchmark at "Xhigh effort." Mythos 5.2 shows a substantial jump in cybersecurity benchmarks, rising from 61.7% to 76.3%. Anthropic identifies Mythos 5.2 as its most proficient model for cybersecurity engineering, adept at identifying and rectifying vulnerabilities.
The new models also demonstrate enhanced capabilities in mathematics, science, and engineering. Additionally, testers of Fable 5.2 found the model more suited to ambiguous tasks requiring abstract reasoning for research and analysis. Anthropic emphasizes the alignment of Fable 5.2 and Mythos 5.2, ensuring the human user's intention is accurately executed by the model.
The company's focus on alignment becomes increasingly crucial as agentic AI becomes more prevalent, as there are growing concerns about AI models potentially veering off course and causing devastating consequences.
Notably, Fable 5.2 includes guardrails to prevent risky cybersecurity activities, such as penetration testing and exploit generation. Anthropic also boasts advanced biology capabilities in Fable 5.2, albeit with stringent safeguards to prevent misuse. Eligible Claude subscribers will receive one to three Fable resets this week, contingent on their subscription duration.
Written by urgent.news from CNET's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
