Mathematical Reasoning on AIME 2025 (pass@16)
85Pass@16Ministral 3
Evaluation Results
| Method | Links | |
|---|---|---|
| Ministral 3Model Size=14B2026.01 | 85 | |
| Qwen3-VLModel Size=8B2026.01 | 79.8 | |
| Ministral 3Model Size=8B2026.01 | 78.7 | |
| Qwen 3Model Size=14B2026.01 | 73.7 | |
| Ministral 3Model Size=3B2026.01 | 72.1 | |
| Qwen3-VLModel Size=4B2026.01 | 69.7 | |
| GEB-1/πBackbone=Qwen2.5-7B2025.09 | 29.48 | |
| GEB-arctanhπBackbone=Qwen2.5-7B2025.09 | 29.38 | |
| GEB-πBackbone=Qwen2.5-7B2025.09 | 28.23 | |
| DPOBackbone=Qwen2.5-7B2025.09 | 23.23 |