🏅 LLM Benchmark
📈 Metrics through time
📝 About
❗FAQ
🚀 Submit
🏅 LLM Benchmark
📈 Metrics through time
📝 About
❗FAQ
🚀 Submit
Textbox
Select columns to show
model_type_symbol
Average ⬆️
ARC-c
ARC-e
Architecture
Boolq
Compute dtype
Double Quant
Group Size
HellaSwag
Lambada
Hub License
Hub ❤️
MMLU
#Size (G)
Openbookqa
#Params (B)
Piqa
Precision
Quant type
Model sha
Truthfulqa
Weight dtype
Winogrande
Model parameters (in billions of parameters)
?
~1.5
~3
~7
~13
~35
~60
70+
Model sizes (GB, int4)
?
~4
~8
~16
~36
~48
~64
>72
Quantization types
🟢 GPTQ
⭐ AQLM
🟩 AWQ
🔷 llama.cpp
🔵 bitsandbytes
🍒 AutoRound
? ?
Weight precision
2bit
3bit
4bit
8bit
16bit
32bit
?
Compute Dtype
Weight Dtype
Double Quant
Group Size
Accuracy Comparison (Selected Models from Table)
Compare
📙 Citation
▼