MATH (Level 5) math Leaderboard
Competition-level math problems, hardest level (5). 4-shot accuracy. · Metric: Accuracy (higher is better)
| # | Model | Accuracy | Paper |
|---|---|---|---|
| 1 | CombinHorizon/huihui-ai-abliterated-Qwen2.5-32B-Inst-BaseMerge-TIES | 59.44 | — |
| 2 | CombinHorizon/zetasepic-abliteratedV2-Qwen2.5-32B-Inst-BaseMerge-TIES | 58.53 | — |
| 3 | BenevolenceMessiah/Qwen2.5-72B-2x-Instruct-TIES-v1.0 | 57.85 | — |
| 4 | Cran-May/tempmotacilla-cinerea-0308 | 55.51 | — |
| 5 | DavidAU/DeepSeek-R1-Distill-Qwen-25.5B-Brainstorm | 55.36 | — |
| 6 | CombinHorizon/huihui-ai-abliteratedV2-Qwen2.5-14B-Inst-BaseMerge-TIES | 54.76 | — |
| 7 | 1024m/QWEN-14B-B100 | 54.38 | — |
| 8 | CombinHorizon/Josiefied-abliteratedV4-Qwen2.5-14B-Inst-BaseMerge-TIES | 53.17 | — |
| 9 | CultriX/Qwen2.5-14B-HyperMarck-dl | 52.87 | — |
| 10 | Azure99/Blossom-V6-14B | 52.57 | — |
| 11 | Daemontatox/CogitoZ | 52.42 | — |
| 12 | CultriX/Qwen2.5-14B-ReasoningMerge | 52.04 | — |
| 13 | Daemontatox/PathFinderAI2.0 | 50.76 | — |
| 14 | Daemontatox/PathFinderAi3.0 | 50.45 | — |
| 15 | CombinHorizon/Rombos-Qwen2.5-7B-Inst-BaseMerge-TIES | 49.32 | — |
| 16 | CultriX/SeQwence-14Bv3 | 47.66 | — |
| 17 | CultriX/SeQwence-14Bv2 | 47.58 | — |
| 18 | Daemontatox/PathfinderAI | 47.58 | — |
| 19 | Daemontatox/mini_Pathfinder | 47.51 | — |
| 20 | Azure99/Blossom-V6-7B | 45.85 | — |
| 21 | EpistemeAI/DeepThinkers-Phi4 | 45.85 | — |
| 22 | CultriX/Qwen2.5-14B-partialmergept1 | 45.39 | — |
| 23 | CultriX/Qwen2.5-14B-Wernicke-SLERP | 44.86 | — |
| 24 | Cran-May/T.E-8.1 | 44.56 | — |
| 25 | Baptiste-HUVELLE-10/LeTriomphant2.2_ECE_iLAB | 44.49 | — |
| 26 | Cran-May/merge_model_20250308_2 | 43.81 | — |
| 27 | CultriX/Qwen2.5-14B-Emergedv3 | 43.58 | — |
| 28 | CultriX/Qwen2.5-14B-Unity | 43.13 | — |
| 29 | EVA-UNIT-01/EVA-Qwen2.5-72B-v0.2 | 43.13 | — |