List your company
ResearchFormerly WizWand

General Knowledge on CMMLU

89.5AccuracyQwen3.5-35B-A3B
13.496833.228452.9672.6916Dec 25, 2025Jan 17, 2026Feb 9, 2026Mar 5, 2026Mar 28, 2026Apr 20, 2026May 14, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Qwen3.5-35B-A3B#Token=6002026.0489.5
Qwen3-Next-80B-A3B#Token=5002026.0489.1
GLM 4.6Evaluation Mode=Chat2025.1288.4
LongCat-Flash Exp-ChatEvaluation Mode=Chat2025.1287.5
DeepSeek V3.2Evaluation Mode=Chat2025.1287.3
Qwen3-30B-A3B#Token=5002026.0487.1
JoyAI-LLM Flash#Token=4002026.0486.7
Qwen3-Next-80B-A3B-Instruct2026.0386.5
Qwen3.5-35B-A3B-BaseEvaluation Framework=OpenCompass, Decoding=Greedy2026.0486.1
Qwen3-Omni-A3B-Instruct2026.0385.76
Qwen3-14B-BaseBackbone=Qwen3-14B-Base, CPT Strategy=None2025.1284.92
LPS-CPTBackbone=Qwen3-14B-Base, CPT Strategy=Low-PPL Sampling2025.1284.79
DOS-CPTBackbone=Qwen3-14B-Base, CPT Strategy=Distance-to-Optimum Selection2025.1284.78
HPS-CPTBackbone=Qwen3-14B-Base, CPT Strategy=High-PPL Sampling2025.1284.53
RS-CPTBackbone=Qwen3-14B-Base, CPT Strategy=Random Data Sampling2025.1284.5
LongCat-Flash ChatEvaluation Mode=Chat2025.1284.3
Qwen3-30B-A3BSparsity Level=Base, K=82026.0583.69
Qwen3-30B-A3B-BaseEvaluation Framework=OpenCompass, Decoding=Greedy2026.0483.6
JoyAI-LLM Flash-BaseEvaluation Framework=OpenCompass, Decoding=Greedy2026.0483.1
LongCat-Next2026.0382.13
BEAMSparsity Level=Mid Sparsity, beta=0.012026.0581.53
GLM-4.7-Flash-T#Token=49002026.0480.2
MoE-DynamicSparsity Level=Mid Sparsity, phi=0.32026.0578.93
Top-K ReducedSparsity Level=Mid Sparsity, K=42026.0578.27
Top-K PruningSparsity Level=Mid Sparsity, K=42026.0576.27
Kimi-Linear-48B-A3B2026.0376.26
Full AttentionSetting=CPT setting2026.0475
BEAMSparsity Level=High Sparsity, beta=0.12026.0574.54
Hybrid-SWASetting=CPT setting2026.0473.69
KSASetting=CPT setting2026.0473.29
Top-K ReducedSparsity Level=High Sparsity, K=22026.0572.99
Hybrid-KSASetting=CPT setting2026.0472.63
Hybrid-SCASetting=CPT setting2026.0472.59
BEAMSparsity Level=Extreme Sparsity, beta=1.02026.0569.36
Hybrid-LinearSetting=CPT setting2026.0468.41
MoE-DynamicSparsity Level=High Sparsity, phi=0.12026.0563.67
AdaMoESparsity Level=Mid Sparsity, Null=1282026.0563.39
HY-1.8B-FP16Model Scale=1.8B, Precision=FP16, Instruction-tuned=true2026.0255.07
Top-K ReducedSparsity Level=Extreme Sparsity, K=12026.0553.05
HY-1.8B-INT4Model Scale=1.8B, Precision=INT4, Instruction-tuned=true2026.0250.8
HY-1.8B-2BitModel Scale=1.8B, Precision=2Bit, Instruction-tuned=true2026.0249.32
Hybrid-GDNsetting=Train-from-scratch2026.0447.19
Hybrid-KSAsetting=Train-from-scratch2026.0446.88
Hybrid-SCAsetting=Train-from-scratch2026.0446.42
Hybrid-SWAsetting=Train-from-scratch2026.0445.89
KSAsetting=Train-from-scratch2026.0445.59
Fullsetting=Train-from-scratch2026.0444.41
AdaMoESparsity Level=High Sparsity, Null=2562026.0537.76
HY-0.5B-FP16Model Scale=0.5B, Precision=FP16, Instruction-tuned=true2026.0237.08
Top-K PruningSparsity Level=High Sparsity, K=22026.0516.42