This screenshot shows the top-20 Open Weight LLMs according to some meta index computed by llm-stats.com¹.

With all due caveats involved in benchmarking LLMs, four points stick out:

1. The current top models of the two well-known US players score 56.8 and 55.8 in the same leaderboard. Open weight models have essentially caught up.

2. Almost all top Open Weight models are currently Chinese.

3. The pace of evolution is mind-boggling. Look at the “Released” column. Whatever you tested three months ago is outdated.

4. Qwen 3.8 27b stands out as way smaller than its peers at 27.8 billion parameters. This thing runs on my notebook and is smarter than me.

---