Anthropic Releases August 2026 Risk Report, Revealing Undisclosed AI Model Model2 Outperforms Claude Mythos 5

On August 15, 2026, AI research company Anthropic released its "August 2026 Risk Report," revealing an unreleased AI model named Model2. According to the report, Model2 outperforms ClaudeMythos5—the currently known strongest model—across multiple tasks. While the performance gains are modest, they primarily reflect enhanced overall capabilities. Citing results from its internal benchmark "CoBenchv2," the report states that Model2 surpasses ClaudeMythos5 in numerous evaluation categories. The report also references Model1, sparking speculation that Model2 may be its successor. Some analysts suggest that both Model1 and Model2 could represent further iterations of ClaudeMythosPreview.