Anthropic releases Claude Fable 5.2 and Claude Mythos 5.2
Original title: Anthropic's Claude Fable 5.2 Beats GPT-6 Cyber in Multi-Stage Cyberattack Benchmark
Anthropic says Fable 5.2 finished 76.3 percent of CyScenarioBench tasks, ahead of GPT-6 Cyber at 61 percent.
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2, citing improvements in coding, reasoning, and vision. It says high-stakes misalignment risk remains low and the models matched or beat earlier Claude models on most alignment measures. On Irregular’s CyScenarioBench, Fable 5.2 completed 76.3 percent of challenges versus 61 percent for GPT-6 Cyber. Mythos 5.2 cyber access remains through the Cyber Verification Program, and Fable 5.2 is available on major clouds.
Unconfirmed: Capability, alignment, and benchmark figures are company or cited-lab claims, not independently verified here.
- Fable 5.2 scored 89.6 percent on Anthropic’s CVE Rediscovery set of 1,829 advisories.
- Anthropic says the models were less likely than Opus 5.5 to publish harmful packages unprompted.
- Claude Security gained continuous code scanning and automatic fix tracking.
- Products
- Claude Security
- Organizations
- Anthropic
From the original article
2 min read at thehackernews.com
Anthropic on Monday released Claude Fable 5. 2 and Claude Mythos 5. 2 models, which it said exhibit improvements in coding, reasoning, and vision.
thehackernews.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.