Open LLM Leaderboard integrates CO₂ emissions for model evaluations

AI Impact Summary

The Open LLM Leaderboard now incorporates carbon emission estimates for model evaluations, revealing significant differences in energy consumption across various models. Notably, community fine-tunes demonstrate superior CO2 efficiency compared to official models, suggesting a potential avenue for reducing the environmental impact of LLM evaluations. This insight highlights the importance of considering resource efficiency alongside model performance when selecting and evaluating language models.