LiveBench's reasoning category — tests logical reasoning, deduction, and analytical thinking with verifiable problems.
Problems use verifiable answers, reducing subjectivity versus human preference votes; monthly refresh helps keep reasoning signal current as models improve.
Coverage is bounded to included reasoning templates and is not exhaustive; category performance can move with monthly task composition rather than only model capability.
Higher is better. Accuracy on reasoning tasks.
Max score: 100
Coverage is bounded to included reasoning templates and is not exhaustive.
| # | Model | Score |
|---|---|---|
| 1 | 99.7 | |
| 2 | 99.2 | |
| 3 | 98.7 | |
| 4 | 95.1 | |
| 5 | 92.7 | |
| 6 | 92.6 | |
| 7 | 92.5 | |
| 8 | 92.4 | |
| 9 | 92.2 | |
| 10 | 92.2 | |
| 11 | 91.7 | |
| 12 | 91.7 | |
| 13 | 91.6 | |
| 14 | 91.6 | |
| 15 | 91.2 | |
| 16 | 91.2 | |
| 17 | 91.2 | |
| 18 | 91.1 | |
| 19 | 90.7 | |
| 20 | 90.6 | |
| 21 | 90.6 | |
| 22 | 90.6 | |
| 23 | 90.5 | |
| 24 | 90.3 | |
| 25 | 90.2 | |
| 26 | 90.1 | |
| 27 | 90.0 | |
| 28 | 89.7 | |
| 29 | 89.7 | |
| 30 | 89.7 | |
| 31 | 89.7 | |
| 32 | 89.6 | |
| 33 | 89.3 | |
| 34 | 89.2 | |
| 35 | 89.1 | |
| 36 | 89.1 | |
| 37 | 88.7 | |
| 38 | 88.7 | |
| 39 | 88.7 | |
| 40 | 88.7 | |
| 41 | 88.6 | |
| 42 | 88.4 | |
| 43 | 88.2 | |
| 44 | 88.2 | |
| 45 | 88.1 | |
| 46 | 88.1 | |
| 47 | 87.8 | |
| 48 | 87.7 | |
| 49 | 87.6 | |
| 50 | 87.6 |