training

162.69 category score

Model	Precision	Sustained
MobileNetV3	FP32	242.26
MobileNetV3	BF16	155.13
Flan-T5 Small	FP32	201.41
Flan-T5 Small	BF16	122.02
DistilGPT2	FP32	216.26
DistilGPT2	BF16	139.11
DistilBERT	FP32	252.88
DistilBERT	BF16	213.29
GraphGPS (Peptides)	FP32	77.55
GraphGPS (Peptides)	BF16	93.72

inference

237.65 category score

Model	Precision	Sustained
MobileNetV3	FP16	775.12
MobileNetV3	INT8	213.32
Flan-T5 Small	FP16	515.98
Flan-T5 Small	INT8	338.76
Flan-T5 Small	FP8	0.00
DistilGPT2	FP16	404.76
DistilGPT2	INT8	235.15
DistilGPT2	FP8	0.00
DistilBERT	FP16	398.20
DistilBERT	INT8	278.09
DistilBERT	FP8	0.00
GraphGPS (Peptides)	FP16	220.48
GraphGPS (Peptides)	INT8	115.37
GraphGPS (Peptides)	FP8	0.00

compute

64.77 category score

Model	Precision	Sustained
Dense MatMul	FP32	1096.32
Dense MatMul	FP16	712.07
Dense MatMul	BF16	737.17
Dense MatMul	FP64	33.39
Sparse MatMul	FP32	7.67
Sparse MatMul	FP64	5.34

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