Another AI model drops, claims it can solve black holes while charging you extra for 'fast mode'

Anthropic has released new versions of its Claude Fable and Claude Mythos AI models, claiming significant improvements in coding, reasoning, and vision capabilities. According to the company's announcement, these models supposedly helped engineers complete complex software migrations, designed electronic devices from scratch, made progress on robotics evaluations, and even formalized a proof of Fermat's Last Theorem—all presented without independent verification or benchmark comparisons.

The real-world impact is that developers now have another subscription-tier AI option with pricing at $8 per million input tokens and $40 per million output tokens, plus an optional 'fast mode' that costs twice as much. While the company claims stronger safeguards and fewer unnecessary blocks, this represents yet another incremental model update in the endless AI arms race, where vendors promise miraculous breakthroughs while quietly raising pricing tiers and introducing premium speed options. The takeaway is familiar: another day, another AI model announcement with grandiose claims but limited transparency about what actually changed under the hood.