Anthropic’s new AI model achieves over 50% on all benchmarks in METR analysis

Anthropic official brand assets (anthropic.com) Anthropic’s latest internal AI model, referred to as Model 4 by METR, has demonstrated notable performance by achieving a score above 50% on every benchmark it was tested on, according to a recent analysis. This development marks the first time in METR’s dataset that a model has cleared the 50% threshold across all tested benchmarks. The model completed its training last month and has quickly become the highest-scoring model in METR’s analysis. This achievement suggests significant advancements in Anthropic’s AI capabilities, reflecting rapid progress compared to previous periods. The news about Model 4’s performance has implications for Anthropic’s standing in the competitive AI landscape. Currently, market participants appear to be adjusting their expectations regarding which company will hold the best AI model by the end of November 2026. Anthropic’s advancements are occurring amid a broader technological race among top AI firms, including Google, Meta, and OpenAI, each vying for dominance in the field. While Anthropic has yet to publicly associate Model 4 with any of its commercial products, the announcement of its Claude Opus 5.5...