METR has revealed details about Anthropic's unreleased Model 4, which was trained between September and October of 2026. This model achieved a score of 178.6 on the AI capability index. This represents a significant leap in performance, nearly doubling the improvement seen between Anthropic's previous models, Opus 4.6 and Mythos 5. AI
IMPACT This benchmark suggests a significant acceleration in AI model capabilities, potentially influencing future development and competitive landscapes.
RANK_REASON The cluster reports on a benchmark score for an unreleased model from a major AI lab, which is a research milestone. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →