Mathematical Reasoning on AIME 2024 (avg@32)
81.35avg@32Qwen3-Next-80B-A3B-Instruct
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3-Next-80B-A3B-InstructArchitecture=MoE, # Total Params=80B, # Activated Params=3B2026.01 | 81.35 | |
| LongCat-Flash-LiteArchitecture=MoE + NE, # Total Params=68.5B, # Activated Params=2.9B~4.5B2026.01 | 72.19 | |
| Kimi-Linear-48B-A3BArchitecture=MoE, # Total Params=48B, # Activated Params=3B2026.01 | 70.52 | |
| Gemini 2.5 Flash-Lite2026.01 | 63.33 | |
| PMD-MEANBackbone=Qwen3-30B-A3B-Base, tau=0.1, Staleness=162026.02 | 50.83 | |
| PMD-MEANBackbone=Qwen3-30B-A3B-Base, tau=0.01, Staleness=162026.02 | 50 | |
| RLBitwidth=BF162026.02 | 40.73 | |
| QuRL w/ UAQBitwidth=INT82026.02 | 40.52 | |
| QuRL w/o UAQBitwidth=INT82026.02 | 39.06 | |
| FlashRLBitwidth=INT82026.02 | 36.77 | |
| GRPOBackbone=Qwen3-30B-A3B-Base, tau=null, Staleness=162026.02 | 36.56 | |
| RLBitwidth=INT82026.02 | 33.95 | |
| BaseBitwidth=BF162026.02 | 28.54 | |
| PMD-MEANBackbone=Qwen2.5-7B, tau=0.02, Staleness=162026.02 | 22.5 | |
| PMD-MEANBackbone=Qwen2.5-7B, tau=0.005, Staleness=162026.02 | 19.69 | |
| On-policyBackbone=Qwen2.5-7B, tau=null, Staleness=12026.02 | 18.65 | |
| PMD-MEANBackbone=Qwen2.5-7B, tau=0.01, Staleness=162026.02 | 17.6 | |
| GRPOBackbone=Qwen2.5-7B, tau=null, Staleness=162026.02 | 17.08 |