Anthropic posted that its second Risk Report is available under the Responsible Scaling Policy. The post links to the full document on anthropic.com. Multiple users quote sections noting two unreleased internal models, Model 1 and Model 2. They state the report describes Claude Mythos 5 and Model 2 as the company's most capable models and the ones used most internally. One post reports Model 2 scores 1.5 points higher than Mythos 5 on Anthropic's internal AECI metric. Another notes the company does not assess Model 2 as speeding AI R&D by more than 2x.