Does that hold when the questions span many subjects instead of one?

1. 1Claude Fable 570.9
2. 2Claude Opus 569.3
3. 3Grok 4.666.8
4. 4GPT-5.6 Sol (max)max reasoning effort66.6
5. 5GPT-5.6 Sol (high)high reasoning effort64.6
6. 6Kimi K359.9
7. 7Muse Spark57.2
8. 8Claude Fable 5.147.3
9. 9Gemini 3.639.7
10. 10TMInkling35.6
11. 11Claude Sonnet 534.6
12. 12GLM 5.220.7
13. 13MiniMax M318.3
14. 14MIMAI Thinking17.5
15. 15Mistral Medium 3.59.2
16. 16Nemotron 3.5 Lightning4.9

Score, percent, higher is better. Select a row for its source.Showing 16 of 16