Claude Fable 5.2 and Claude Mythos 5.2 have been released, demonstrating improvements in coding, reasoning, and vision capabilities, with the models outperforming Opus 5.5 in most alignment measures. They can be used for ambitious projects in various fields, and their performance has been validated by external evaluators and testers. AI summary
Firehose
Filtered to Hacker News, tagged “programming” · clear filters
Browse: People · Companies · Papers · Podcasts · Hacker News · Deep dives
Browse by tag
Claude Fable 5.2 and Mythos 5.2 models show improvements in areas like coding, reasoning, and vision, achieving up to 49.1% and 83.5% accuracy on benchmark tests, respectively, at a lower price point ($8 per million input tokens and $40 per million output tokens) compared to previous models. AI summary