Anthropic disclosed Model 2: The performance is slightly higher than the most publicly available Claude Mythos 5 Anthropic In the "August 2026 Risk Report" released on August 15, an unreleased AI model Model 2 was accidentally revealed

Anthropic disclosed Model 2: The performance is slightly higher than the most publicly available Claude Mythos 5 Anthropic In the "August 2026 Risk Report" released on August 15, an unreleased AI model Model 2 was accidentally revealed. The report states that the model outperforms the currently strongest publicly available model, Claude Mythos 5, on multiple tasks. The increase is not large, focusing on the intergenerational polishing of comprehensive capabilities. According to disclosed details, compared with Claude Mythos 5, Model 2 has significantly stronger performance in multiple tasks, but the improvement is relatively limited, mainly reflected in the overall enhancement of AI comprehensive capabilities. The report also comes with the comparison results of the Anthropic internal benchmark "CoBench v2" as evidence. The name "Model 1" also appeared in the report, leading users to speculate that Model 2 may be a subsequent version of Model 1, and both may be further iterations of the Claude Mythos Preview. If true, Anthropic is quietly advancing up the capability ladder with an internal serial number, and the public naming lags behind the actual progress. via AI News (author: AI Base)