Sapient researchers trained a 1B reasoning model on just 40B tokens — scoring competitively with 2B-7B models at a fraction of typical pretraining cost.

The broad safeguards built into Anthropic's Claude Fable 5, a Mythos-class AI model, blocks some mundane requests on cybersecurity and biology.

Most AI models are designed to be autoregressive—they generate text left to right one token at a time. DiffusionGemma has more in common with image generation models, which start with static and then denoise it to create the desired content.

Anthropic made Claude Fable 5 generally available at twice the Opus 4.8 rate, then set its free subscription access to expire June 23. The scarce input is now capacity.