Leaderboard empirical score
Stored empirical_score from the leaderboard. 141 of 10,858 models store a score. The highest is 67 for DeepSeek-V4.1-Flash. Grok 4.5 stores 64. MiMo-V2.6-Pro stores 0, so that bar has no length. A longer bar is a higher stored score. This is the leaderboard empirical score, not the telemetry dyno awarded score and not the stored empirical 100. A model with no stored empirical score stays off the page.
Loading the leaderboard.