Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Language Modeling on FineWeb-Edu 300M×5B (val)
Loading...
23.6
PPL
FP16 reference
-44.536
415.382
875.3
1,335.218
Apr 20, 2026
PPL
Updated 1mo ago
Evaluation Results
Method
Method
Links
PPL
FP16 reference
k=0, Quantization=FP16
2026.04
23.6
FP16 reference
k=64, Quantization=FP1...
2026.04
23.7
FP16 reference
k=64, Quantization=FP1...
2026.04
23.7
QuaRot (per-Linear) W4A4
k=64, Quantization=W4A...
2026.04
25.4
QuaRot (per-Linear) W4A4
k=64, Quantization=W4A...
2026.04
25.4
QuaRot (per-Linear) W4A4
k=0, Quantization=W4A4
2026.04
25.5
QuaRot (full) W4A4
k=0, Quantization=W4A4
2026.04
26
QuaRot (full) W4A4
k=64, Quantization=W4A...
2026.04
26
QuaRot (full) W4A4
k=64, Quantization=W4A...
2026.04
26.1
SmoothQuant W4A4
k=64, Quantization=W4A...
2026.04
39.9
SmoothQuant W4A4
k=0, Quantization=W4A4
2026.04
57.6
SmoothQuant W4A4
k=64, Quantization=W4A...
2026.04
58.1
Naive RTN W4A4
k=64, Quantization=W4A...
2026.04
119
Naive RTN W4A4
k=64, Quantization=W4A...
2026.04
671
Naive RTN W4A4
k=0, Quantization=W4A4
2026.04
1,727
Feedback
Search any
task
Search any
task