training

798.77 category score

Model	Precision	Sustained
MobileNetV3	FP32	602.18
MobileNetV3	BF16	263.18
Flan-T5 Small	FP32	1459.48
Flan-T5 Small	BF16	793.25
DistilGPT2	FP32	1358.73
DistilGPT2	BF16	736.71
DistilBERT	FP32	1677.92
DistilBERT	BF16	1506.93
GraphGPS (Peptides)	FP32	426.96
GraphGPS (Peptides)	BF16	372.84

inference

1899.59 category score

Model	Precision	Sustained
MobileNetV3	FP16	3178.18
MobileNetV3	INT8	1156.30
Flan-T5 Small	FP16	4434.11
Flan-T5 Small	INT8	1521.28
Flan-T5 Small	FP8	1915.20
DistilGPT2	FP16	4070.93
DistilGPT2	INT8	1587.21
DistilGPT2	FP8	2059.57
DistilBERT	FP16	2839.56
DistilBERT	INT8	1155.28
DistilBERT	FP8	1465.34
GraphGPS (Peptides)	FP16	1381.98
GraphGPS (Peptides)	INT8	674.71
GraphGPS (Peptides)	FP8	730.53

compute

464.68 category score

Model	Precision	Sustained
Dense MatMul	FP32	10180.04
Dense MatMul	FP16	12569.96
Dense MatMul	BF16	13113.37
Dense MatMul	FP64	4004.05
Sparse MatMul	FP32	15.61
Sparse MatMul	FP64	27.72

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