List your company

Reasoning on GPQA Diamond

91.9AccuracyGemini-3.0 Pro
43.841656.318368.79581.2717Jul 14, 2025Aug 27, 2025Oct 11, 2025Nov 25, 2025Jan 9, 2026Feb 23, 2026Apr 9, 2026

Evaluation results

MethodLinks
Gemini-3.0 ProVariant=Pro2026.0191.9———————————
GPT-5 HighVariant=High2026.0185.7———————————
Qwen3.5-27BArchitecture=Dense, # Total Params=27B, # Activated Params=27B, Reasoning Mode=REASONING2026.0485.5———————————
RSAModel 2=GPT-5 mini, Pairing Type=Heterogeneous2026.0485—————————0.521
Kimi-K2 ThinkingThinking Mode=true2026.0184.5———————————
MiMo-V2-FlashVariant=Flash2026.0184.3———————————
SQUEEZE EVOLVEModel 1=Qwen3-30B-A3B-I, Model 2=GPT-5 mini, p-value=20, Pairing Type=Heterogeneous2026.0483.6—————————0.351.5
Claude Sonnet 4.5Variant=Sonnet 4.52026.0183.4———————————
DeepSeek-V3.2 ThinkingThinking Mode=true2026.0182.4———————————
PETS-On.LLM=GPT-120B, Confidence-Weighted=false2026.0282.4———————4,172100——
GPT-5 miniReasoning Mode=REASONING: HIGH2026.0482.3———————————
UniformLLM=GPT-120B, Confidence-Weighted=false2026.0281.8———————8,10998——
PETS-On. (conf)LLM=GPT-120B, Confidence-Weighted=true2026.0281.6———————5,09499——
Uniform (conf)LLM=GPT-120B, Confidence-Weighted=true2026.0281.4———————9,59397——
EXAONE 4.5 33BArchitecture=Dense, # Total Params=33B, # Activated Params=32B, Reasoning Mode=REASONING2026.0480.5———————————
K-EXAONE-236B-A23BArchitecture=MoE, # Total Params=236B, # Activated Params=23B, Reasoning Mode=REASONING2026.0479.1———————————
DeepSeek-R1-0528Quantization=FP8-Block-Wise2026.0278.28———————————
DeepSeek-R1-0528Quantization=W4A8-FP82026.0277.37———————————
Qwen3-VL-235B-A22BArchitecture=MoE, # Total Params=236B, # Activated Params=22B, Reasoning Mode=Thinking2026.0477.1———————————
PETS-On. (conf)LLM=Qwen-Long, Confidence-Weighted=true2026.0276.9———————4,146100——
Uniform (conf)LLM=Qwen-Long, Confidence-Weighted=true2026.0276.4———————9,52697——
PETS-On.LLM=Qwen-Long, Confidence-Weighted=false2026.0276.3———————4,852100——
Uniform (conf)LLM=GPT-20B, Confidence-Weighted=true2026.0276.3———————9,89598——
PETS-On. (conf)LLM=GPT-20B, Confidence-Weighted=true2026.0276.1———————7,26999——
UniformLLM=GPT-20B, Confidence-Weighted=false2026.0276.1———————9,69998——
UniformLLM=Qwen-Long, Confidence-Weighted=false2026.0276———————9,19597——
PETS-On.LLM=GPT-20B, Confidence-Weighted=false2026.0275.9———————6,826100——
SQUEEZE EVOLVEModel 1=Qwen3-30B-A3B-I, Model 2=Qwen3-30B-A3B-T, p-value=0, Pairing Type=Homogeneous2026.0475.9—————————0.321.8
RSAModel 2=Qwen3-30B-A3B-T, Pairing Type=Homogeneous2026.0474—————————0.571
Uniform (conf)LLM=Qwen3-30B, Confidence-Weighted=true2026.0272———————9,25796——
PETS-On. (conf)LLM=Qwen3-30B, Confidence-Weighted=true2026.0271.8———————4,35799——
PETS-On.LLM=Qwen3-30B, Confidence-Weighted=false2026.0271.6———————3,826100——
UniformLLM=Qwen3-30B, Confidence-Weighted=false2026.0271.2———————8,45697——
PETS-On. (conf)LLM=Qwen3-4B, Confidence-Weighted=true2026.0269.6———————4,888100——
Uniform (conf)LLM=Qwen3-4B, Confidence-Weighted=true2026.0269.4———————9,85697——
PETS-On.LLM=Qwen3-4B, Confidence-Weighted=false2026.0268.8———————4,662100——
PIRBackbone=DeepSeek-R1-32B2026.0268.6——————3,048.82————
UniformLLM=Qwen3-4B, Confidence-Weighted=false2026.0268.1———————9,52097——
GPT-4.1Access=Close-source2025.0767.17———————————
GPT-o3-miniAccess=Close-source2025.0766.67———————————
Qwen3-32BAccess=Open-source, Parameters=32B2025.0765.15———————————
Open-source Upper BoundCategory=Baselines2025.0765.15———————————
SMCSCategory=Ours2025.0765.15———————————
Self-MoACategory=Other Methods2025.0764.41———————————
QTIPBackbone=Qwen3-14B, Type=vector, Zero-shot=true2025.1164———————————
Claude-3.7-SonnetAccess=Close-source2025.0763.64———————————
Self Consistency (Best on Validation)Category=Other Methods2025.0763.64———————————
ParoQuantBackbone=Qwen3-14B, Type=linear, Zero-shot=true2025.1163.5———————————
Symbolic-MoE*Category=Other Methods2025.0762.63———————————
FP16Backbone=Qwen3-14B, Zero-shot=true2025.1162.5———————————
AWQBackbone=Qwen3-14B, Type=linear, Zero-shot=true2025.1162———————————
Claude-3.5-SonnetAccess=Close-source2025.0761.62———————————
FP16Backbone=Qwen3-8B, Zero-shot=true2025.1160.3———————————
AWQBackbone=Qwen3-8B, Type=linear, Zero-shot=true2025.1160.2———————————
DeepSeek-R1-Distill-Llama-70BAccess=Open-source, Parameters=70B2025.0760.1———————————
DeepSeek-R1-Distill-Qwen-32BAccess=Open-source, Parameters=32B2025.0760.1———————————
Simple RouterCategory=Other Methods2025.0760.1———————————
GLM-Z1-32B-0414Access=Open-source, Parameters=32B2025.0759.6———————————
QTIPBackbone=Qwen3-8B, Type=vector, Zero-shot=true2025.1159.3———————————
MoACategory=Other Methods2025.0758.8———————————
EXAONE-Deep-32BAccess=Open-source, Parameters=32B2025.0758.59———————————
E-QATBackbone=Qwen3-14B, Type=linear, Zero-shot=true2025.1158.4———————————
ParoQuantBackbone=Qwen3-8B, Type=linear, Zero-shot=true2025.1157.7———————————
QwQ-32BAccess=Open-source, Parameters=32B2025.0757.07———————————
ASPO-Math-8BBackbone=Qwen3-8B2025.1056.3———————————
E-QATBackbone=Qwen3-8B, Type=linear, Zero-shot=true2025.1156.1———————————
DeepSeek-R1-32BBackbone=DeepSeek-R1-32B2026.0255.81——————5,307.51————
PIRBackbone=QwQ-32B2026.0255.56——————4,033.46————
QTIPBackbone=Qwen3-4B, Type=vector, Zero-shot=true2025.1155.2———————————
ParoQuantBackbone=Qwen3-4B, Type=linear, Zero-shot=true2025.1153.7———————————
High Conf.Backbone=DS-R1-Distill-Llama-8B2025.1053———————————
GPT-4oAccess=Close-source2025.0752.53———————————
AWQBackbone=Qwen3-4B, Type=linear, Zero-shot=true2025.1152.2———————————
Loc. uniBackbone=DS-R1-Distill-Llama-8B2025.1052———————————
E-QATBackbone=Qwen3-4B, Type=linear, Zero-shot=true2025.1151———————————
Self-Cert.Backbone=DS-R1-Distill-Llama-8B2025.1051———————————
Self-Cert.Backbone=DS-R1-Distill-Qwen-7B2025.1051———————————
Loc. uniBackbone=DS-R1-Distill-Qwen-7B2025.1051———————————
Gemma-3-27b-itAccess=Open-source, Parameters=27B2025.0750.51———————————
FP16Backbone=Qwen3-4B, Zero-shot=true2025.1150———————————
Glob. non-uniBackbone=DS-R1-Distill-Qwen-7B2025.1050———————————
HuatuoGPT-o1-72BAccess=Open-source, Parameters=72B2025.0750———————————
Low Ent.Backbone=DS-R1-Distill-Llama-8B2025.1049———————————
QwQ-32BBackbone=QwQ-32B2026.0248.84——————6,132.83————
Llama-3.3-Nemotron-Super-49B-v1Access=Open-source, Parameters=49B2025.0748.48———————————
Mean Acc.Backbone=DS-R1-Distill-Llama-8B2025.1048———————————
Mean Acc.Backbone=DS-R1-Distill-Qwen-7B2025.1048———————————
Low Ent.Backbone=DS-R1-Distill-Qwen-7B2025.1048———————————
ParoQuantBackbone=R1-Distill-Llama-8B, Type=linear, Zero-shot=true2025.1147.5———————————
ReASC (online)Model=GEMMA-3-27B2026.0147.21———161.680.29——————
Glob. non-uniBackbone=DS-R1-Distill-Llama-8B2025.1047———————————
Llama-3.3-70B-InstructAccess=Open-source, Parameters=70B2025.0746.97———————————
FP16Backbone=R1-Distill-Llama-8B, Zero-shot=true2025.1146.6———————————
PreSafebase_model=DeepSeek-R1-Distill-Qwen-7B2026.0346.3———————————
SafeChainbase_model=DeepSeek-R1-Distill-Qwen-7B2026.0346.1———————————
High Conf.Backbone=DS-R1-Distill-Qwen-7B2025.1046———————————
Glob. uniBackbone=DS-R1-Distill-Qwen-7B2025.1046———————————
PIRBackbone=DeepSeek-R1-14B2026.0245.83——————3,576.72————
SCModel=GEMMA-3-27B, k=162026.0145.69———368.970.12——————
ESCModel=GEMMA-3-27B, w=42026.0145.69———210.780.22——————