List your company
ResearchFormerly WizWand

Graduate-Level Reasoning on GPQA

78.9AccuracyM2CL
17.85233.70149.5565.399Jul 15, 2024Oct 31, 2024Feb 17, 2025Jun 5, 2025Sep 22, 2025Jan 8, 2026Apr 27, 2026
Updated 6mo ago

Evaluation results

MethodLinks
M2CLBackbone Model=Qwen-72B, Number of LLMs=42026.0278.9
Qwen3-OmniOpen-Source=✓, Size=30B-A3B, no tools=true2026.0473.1
Nemotron 3 Nano 30B-A3BOpen-Source=✓, Size=30B-A3B, no tools=true2026.0473
Nemotron 3 Nano OmniOpen-Source=✓, Size=30B-A3B, no tools=true2026.0472.2
M2CLBackbone Model=Qwen-14B, Number of LLMs=42026.0266.2
M2CLBackbone Model=Qwen-7B, Number of LLMs=42026.0266.1
BLASSTBackbone=Qwen3-8B, Attention Method=BLASST, Sparsity=~50%, Phase=Decode2025.1261.51
Dense AttentionBackbone=Qwen3-8B, Attention Method=Dense, Phase=Decode2025.1261.21
RocketKVBackbone=Qwen3-8B, Attention Method=RocketKV, Phase=Decode2025.1260.5
QuestBackbone=Qwen3-8B, Attention Method=Quest, Phase=Decode2025.1260.12
HCP-MADBase Model=Qwen2.5-32b-instruct + Llama-3.1-70b-instruct2026.0454.04
GPTSwarmBackbone Model=Qwen-72B, Number of LLMs=42026.0252.4
DyLANBackbone Model=Qwen-72B, Number of LLMs=42026.0251.6
DOWNBase Model=Qwen2.5-32b-instruct + Llama-3.1-70b-instruct2026.0451.01
Heter-MADBase Model=Qwen2.5-32b-instruct + Llama-3.1-70b-instruct2026.0450.5
MADBase Model=Llama-3.1-70b-instruct2026.0449.49
MARSBase Model=Qwen2.5-32b-instruct + Llama-3.1-70b-instruct2026.0449.49
CoTBase Model=Llama-3.1-70b-instruct2026.0448.99
SRBase Model=Llama-3.1-70b-instruct2026.0447.98
SCBase Model=Llama-3.1-70b-instruct2026.0446.97
MacNetBackbone Model=Qwen-72B, Number of LLMs=42026.0246.2
MADBase Model=Qwen2.5-32b-instruct2026.0445.96
Best-of-N (BoN)Backbone Model=Qwen-72B, Number of LLMs=42026.0245.9
DebateBackbone Model=Qwen-72B, Number of LLMs=42026.0243.4
CoTBase Model=Qwen2.5-32b-instruct2026.0442.93
SCBase Model=Qwen2.5-32b-instruct2026.0441.92
SRBase Model=Qwen2.5-32b-instruct2026.0441.41
DyLANBackbone Model=Qwen-14B, Number of LLMs=42026.0239.4
GPTSwarmBackbone Model=Qwen-14B, Number of LLMs=42026.0238.7
Best-of-N (BoN)Backbone Model=Qwen-7B, Number of LLMs=42026.0236.4
GPTSwarmBackbone Model=Qwen-7B, Number of LLMs=42026.0235.6
DyLANBackbone Model=Qwen-7B, Number of LLMs=42026.0235.4
Single executionBackbone Model=Qwen-72B, Number of LLMs=42026.0234.9
Qwen2-57B-A14BArchitecture=MoE, # Act Params=14B, # Params=57B2024.0734.3
Best-of-N (BoN)Backbone Model=Qwen-14B, Number of LLMs=42026.0232.8
MacNetBackbone Model=Qwen-14B, Number of LLMs=42026.0231.9
Yi-1.5-34BArchitecture=Dense, # Act Params=32B, # Params=32B2024.0730.8
Qwen1.5-32BArchitecture=Dense, # Act Params=34B, # Params=34B2024.0730.8
MacNetBackbone Model=Qwen-7B, Number of LLMs=42026.0230.8
DebateBackbone Model=Qwen-14B, Number of LLMs=42026.0230.3
Mixtral-8x7BArchitecture=MoE, # Act Params=12B, # Params=47B2024.0729.2
DebateBackbone Model=Qwen-7B, Number of LLMs=42026.0228.7
Single executionBackbone Model=Qwen-14B, Number of LLMs=42026.0221.2
Single executionBackbone Model=Qwen-7B, Number of LLMs=42026.0220.2