How far do open weights trail the frontier?
A LessWrong post explores the performance gap between open-weight and frontier AI models, proposing the Epoch AI Contribution Index (ECI) as a superior metric to simpler indices. The author presents graphs illustrating the divergence in performance over time, noting that GLM-5.2 has not yet been scored. The analysis also touches upon the competitive landscape between major AI labs like OpenAI and Anthropic. AI
IMPACT Analysis suggests open-weight models are falling behind frontier models, highlighting the need for better evaluation metrics.