Hugging Face has released version 2 of the Open LLM Leaderboard, ranking open source language models. The top model, Qwen2-72B-Instruct by Alibaba, leads the board. Chinese AI models dominate the leaderboard, with Qwen’s models occupying three of the top 10 spots. The evaluation criteria include intelligence, reasoning, mathematics ability, and following human instructions. Over 7,500 models were assessed, with Qwen2-72B-Instruct scoring the highest. The results can be viewed on the Hugging Face website. The breakdown of the top 10 models is provided, showing Qwen’s models’ strength. Smaug-72B, which ranked 9th, was the top model in version 1 of the leaderboard. It outperformed GPT-3.5 in several benchmarks. The competition among language models continues to evolve, with new advancements and rankings shaping the field.
Source link
Source link: https://gigazine.net/gsc_news/en/20240701-open-llm-leaderboard-v2/
GIPHY App Key not set. Please check settings