List your company
ResearchFormerly WizWand

Multi-Domain Knowledge on MMLU

80.75MMLU Multi-Domain Knowledge AccFP16
20.814836.374951.93567.4951Jan 13, 2026Feb 6, 2026Mar 3, 2026Mar 28, 2026Apr 22, 2026May 17, 2026Jun 11, 2026
Updated 4mo ago

Evaluation results

MethodLinks
FP16Model=Qwen3-32B-Instruct, #Bits(W)=16, #Bits(A)=162026.0580.75—
FP16Model=Qwen3-32B-Instruct, Weight bits=16, Activation bits=162026.0680.75—
BF16Model=Qwen2.5-14B, A-W Quant Type=BF162026.0280.17—
HiF4+HiGPTQModel=Qwen2.5-14B, A-W Quant Type=HiF4+HiGPTQ2026.0278.65-1.52
NVFP4+PTSModel=Qwen2.5-14B, A-W Quant Type=NVFP4+PTS2026.0278.63-1.54
HiF4Model=Qwen2.5-14B, A-W Quant Type=HiF42026.0278.62-1.55
NVFP4Model=Qwen2.5-14B, A-W Quant Type=NVFP42026.0276.53-3.64
TWLAModel=Qwen3-32B-Instruct, Weight bits=1.58, Activation bits=162026.0675.59—
GPTQModel=Qwen3-32B-Instruct, Weight bits=3, Activation bits=162026.0673.97—
TWLAModel=Qwen3-32B-Instruct, Weight bits=1.58, Activation bits=4MP2026.0670.21—
BWLAModel=Qwen3-32B-Instruct, #Bits(W)=1.15, #Bits(A)=162026.0567.74—
BWLAModel=Qwen3-32B-Instruct, #Bits(W)=1.15, #Bits(A)=62026.0567.48—
BF16Model=Llama3-8B, A-W Quant Type=BF162026.0266.55—
GPTQModel=Qwen3-32B-Instruct, #Bits(W)=3, #Bits(A)=162026.0563.75—
HiF4+HiGPTQModel=Llama3-8B, A-W Quant Type=HiF4+HiGPTQ2026.0263.55-3
HiF4Model=Llama3-8B, A-W Quant Type=HiF42026.0263.37-3.18
BF16Model=Mistral-7B, A-W Quant Type=BF162026.0263.3—
NVFP4+PTSModel=Llama3-8B, A-W Quant Type=NVFP4+PTS2026.0262.81-3.74
ARB-LLMModel=Qwen3-32B-Instruct, #Bits(W)=1.06, #Bits(A)=162026.0562.56—
NVFP4Model=Llama3-8B, A-W Quant Type=NVFP42026.0261.64-4.91
HiF4Model=Mistral-7B, A-W Quant Type=HiF42026.0261.63-1.67
HiF4+HiGPTQModel=Mistral-7B, A-W Quant Type=HiF4+HiGPTQ2026.0261.53-1.77
NVFP4+PTSModel=Mistral-7B, A-W Quant Type=NVFP4+PTS2026.0261.15-2.15
T-SPINIteration=Iter42026.0158.55—
T-SPINIteration=Iter32026.0158.51—
SFTMethod=Supervised Fine-Tuning2026.0158.37—
SPINIteration=Iter42026.0158.05—
SPINIteration=Iter02026.0157.97—
Mistral-7BConfiguration=Base2026.0157.86—
SPINIteration=Iter32026.0157.8—
T-SPINIteration=Iter02026.0157.74—
SPINIteration=Iter22026.0157.74—
T-SPINIteration=Iter42026.0157.68—
T-SPINIteration=Iter32026.0157.67—
SPINIteration=Iter12026.0157.63—
T-SPINIteration=Iter22026.0157.61—
SPINIteration=Iter22026.0157.49—
T-SPINIteration=Iter12026.0157.48—
SFT2026.0157.29—
Zephyr-7B2026.0156.9—
T-SPINIteration=Iter12026.0156.89—
T-SPINIteration=Iter22026.0156.89—
SPINIteration=Iter12026.0156.86—
T-SPINIteration=Iter02026.0156.42—
SPINIteration=Iter02026.0156.25—
SPINIteration=Iter32026.0155.88—
SliM-LLMModel=Qwen3-32B-Instruct, Weight bits=2MP, Activation bits=162026.0655.38—
SPINIteration=Iter42026.0153.59—
BiLLMModel=Qwen3-32B-Instruct, #Bits(W)=1.06, #Bits(A)=162026.0549.04—
PB-LLMModel=Qwen3-32B-Instruct, Weight bits=1.7, Activation bits=162026.0648.21—
BF16Model=Llama2-7B, A-W Quant Type=BF162026.0246.52—
HiF4Model=Llama2-7B, A-W Quant Type=HiF42026.0244.54-1.98
HiF4+HiGPTQModel=Llama2-7B, A-W Quant Type=HiF4+HiGPTQ2026.0243.89-2.63
NVFP4+PTSModel=Llama2-7B, A-W Quant Type=NVFP4+PTS2026.0243.43-3.09
NVFP4Model=Llama2-7B, A-W Quant Type=NVFP42026.0241.51-5.01
GPTQ + QIGModel=LLaMA-2-7B, Bit-width=3-bit2026.0332.01—
GPTQModel=LLaMA-2-7B, Bit-width=3-bit2026.0330.05—
GPTQModel=Qwen3-32B-Instruct, Weight bits=2, Activation bits=162026.0627.1—
NVFP4Model=Mistral-7B, A-W Quant Type=NVFP42026.0226.79-36.51
BiLLMModel=Qwen3-32B-Instruct, #Bits(W)=1.06, #Bits(A)=62026.0525.82—
GPTQModel=Qwen3-32B-Instruct, #Bits(W)=3, #Bits(A)=62026.0525.36—
GPTQModel=Qwen3-32B-Instruct, #Bits(W)=2, #Bits(A)=62026.0524.75—
GPTQModel=Qwen3-32B-Instruct, #Bits(W)=2, #Bits(A)=162026.0524.58—
ARB-LLMModel=Qwen3-32B-Instruct, #Bits(W)=1.06, #Bits(A)=62026.0524.33—
SliM-LLMModel=Qwen3-32B-Instruct, Weight bits=2MP, Activation bits=42026.0623.71—
GPTQModel=Qwen3-32B-Instruct, Weight bits=3, Activation bits=42026.0623.28—
GPTQModel=Qwen3-32B-Instruct, Weight bits=2, Activation bits=42026.0623.2—
PB-LLMModel=Qwen3-32B-Instruct, Weight bits=1.7, Activation bits=42026.0623.12—