METR says Anthropic’s Model 4 is the first to clear 50% on every benchmark

Anthropic’s Model 4 has become the first AI model to score above 50% on every benchmark evaluated by METR, the nonprofit known for assessing frontier-model capabilities. The milestone arrives as METR warns that static benchmark suites are struggling to keep pace with rapid model gains, arguing for adaptive evaluation frameworks that can be refreshed as systems improve. The result matters because benchmark saturation can obscure meaningful differences between leading models and complicate how developers, enterprises and investors judge progress. What to watch next is whether METR and peers publish replacement or harder evaluations, and whether Anthropic discloses fuller Model 4 results and safety testing.