training

583.16 category score

Model	Precision	Sustained
MobileNetV3	FP32	548.88
MobileNetV3	BF16	302.88
Flan-T5 Small	FP32	904.71
Flan-T5 Small	BF16	494.04
DistilGPT2	FP32	867.54
DistilGPT2	BF16	451.21
DistilBERT	FP32	1212.77
DistilBERT	BF16	861.84
GraphGPS (Peptides)	FP32	325.28
GraphGPS (Peptides)	BF16	336.85

inference

1177.35 category score

Model	Precision	Sustained
MobileNetV3	FP16	2973.56
MobileNetV3	INT8	731.67
Flan-T5 Small	FP16	2611.39
Flan-T5 Small	INT8	1093.17
Flan-T5 Small	FP8	1263.89
DistilGPT2	FP16	1733.83
DistilGPT2	INT8	953.43
DistilGPT2	FP8	976.68
DistilBERT	FP16	2078.75
DistilBERT	INT8	1153.13
DistilBERT	FP8	1380.25
GraphGPS (Peptides)	FP16	594.99
GraphGPS (Peptides)	INT8	287.57
GraphGPS (Peptides)	FP8	295.47

compute

362.95 category score

Model	Precision	Sustained
Dense MatMul	FP32	7444.14
Dense MatMul	FP16	7466.03
Dense MatMul	BF16	7641.46
Dense MatMul	FP64	113.08
Sparse MatMul	FP32	26.68
Sparse MatMul	FP64	19.81

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