Voltropy unveiled Vast-10M, a new model family with a 10-million-token native context window, achieved through a novel Voltropy Scalable Attention (VSA) algorithm. Vast-10M-Flash outperforms Claude Fable 5.1 and matches GPT-6 Astra within their context windows while maintaining performance at 10× larger contexts, enabling analysis of massive datasets like the entire U.S. tax code or full codebases.