training

383.74 category score

Model	Precision	Sustained
MobileNetV3	FP32	346.28
MobileNetV3	BF16	173.31
Flan-T5 Small	FP32	579.06
Flan-T5 Small	BF16	336.21
DistilGPT2	FP32	609.72
DistilGPT2	BF16	343.98
DistilBERT	FP32	736.08
DistilBERT	BF16	639.53
GraphGPS (Peptides)	FP32	212.64
GraphGPS (Peptides)	BF16	214.16

inference

578.03 category score

Model	Precision	Sustained
MobileNetV3	FP16	1472.07
MobileNetV3	INT8	459.72
Flan-T5 Small	FP16	1411.06
Flan-T5 Small	INT8	672.11
Flan-T5 Small	FP8	0.00
DistilGPT2	FP16	1390.64
DistilGPT2	INT8	630.82
DistilGPT2	FP8	0.00
DistilBERT	FP16	1102.80
DistilBERT	INT8	531.81
DistilBERT	FP8	0.00
GraphGPS (Peptides)	FP16	525.67
GraphGPS (Peptides)	INT8	260.48
GraphGPS (Peptides)	FP8	0.00

compute

199.17 category score

Model	Precision	Sustained
Dense MatMul	FP32	3178.71
Dense MatMul	FP16	3849.72
Dense MatMul	BF16	4187.37
Dense MatMul	FP64	1178.44
Sparse MatMul	FP32	9.22
Sparse MatMul	FP64	16.39

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