training

36.94 category score

Model	Precision	Sustained
MobileNetV3	FP32	39.36
MobileNetV3	BF16	27.32
Flan-T5 Small	FP32	24.73
Flan-T5 Small	BF16	75.82
DistilGPT2	FP32	13.76
DistilGPT2	BF16	73.09
DistilBERT	FP32	15.62
DistilBERT	BF16	96.57
GraphGPS (Peptides)	FP32	16.83
GraphGPS (Peptides)	BF16	16.85

inference

119.10 category score

Model	Precision	Sustained
MobileNetV3	FP16	83.49
MobileNetV3	INT8	32.28
Flan-T5 Small	FP16	431.92
Flan-T5 Small	INT8	164.70
Flan-T5 Small	FP8	198.92
DistilGPT2	FP16	283.05
DistilGPT2	INT8	148.71
DistilGPT2	FP8	153.54
DistilBERT	FP16	241.51
DistilBERT	INT8	130.60
DistilBERT	FP8	131.80
GraphGPS (Peptides)	FP16	72.75
GraphGPS (Peptides)	INT8	34.33
GraphGPS (Peptides)	FP8	35.85

compute

110.36 category score

Model	Precision	Sustained
Dense MatMul	FP32	74.12
Dense MatMul	FP16	1202.31
Dense MatMul	BF16	1203.47
Dense MatMul	FP64	27.81
Sparse MatMul	FP32	21.07
Sparse MatMul	FP64	17.79

Provisional results — not the final, official release · GT = LynxBenchAI Global Score. Every precision result is sustained throughput under a declared, bounded optimisation budget — no collapsed single number below the category level.

← Back to the leaderboard

LynxBenchAI End-User License Agreement · LynxBenchAI Privacy Policy