Anthropic says its new model can solve black hole physics puzzles; what could go wrong?
Anthropic’s Says Its New AI Model Can Solve Black Hole Mysteries. What Could Go Wrong?
Anthropic released Claude Fable 5.2 and Mythos 5.2, saying the two models solved two previously unsolved black hole physics problems and set the price at $8 per million input tokens and $40 per million output tokens. The company says output speed is up to 30% faster than Fable 5.1 and Mythos 5.1, and also says about 1,000 Fable 5.2 sub-agents converted the existing proof of Fermat's Last Theorem into a computer-verifiable form in 30 hours, whereas previous versions of Claude needed 11 days; the two black hole research papers have not yet undergone independent peer review. Anthropic says alignment has improved in 5.2 but acknowledges that both models suspect they are being evaluated more often than the previous generation.
In addition to outlining the capabilities and pricing of the two new models, the report mentions that they more often suspect they are being evaluated than previous generations, which may be useful for readers interested in evaluation awareness.