training

1057.55 category score

Model	Precision	Sustained
MobileNetV3	FP32	642.84
MobileNetV3	BF16	253.32
Flan-T5 Small	FP32	2203.37
Flan-T5 Small	BF16	1099.16
DistilGPT2	FP32	1928.26
DistilGPT2	BF16	703.68
DistilBERT	FP32	2448.42
DistilBERT	BF16	2127.79
GraphGPS (Peptides)	FP32	612.45
GraphGPS (Peptides)	BF16	575.06

inference

2085.33 category score

Model	Precision	Sustained
MobileNetV3	FP16	4720.15
MobileNetV3	INT8	1605.82
Flan-T5 Small	FP16	6991.61
Flan-T5 Small	INT8	826.52
Flan-T5 Small	FP8	2362.98
DistilGPT2	FP16	1900.43
DistilGPT2	INT8	563.79
DistilGPT2	FP8	906.52
DistilBERT	FP16	4366.17
DistilBERT	INT8	733.72
DistilBERT	FP8	1863.41
GraphGPS (Peptides)	FP16	2274.51
GraphGPS (Peptides)	INT8	793.56
GraphGPS (Peptides)	FP8	1157.93

compute

551.76 category score

Model	Precision	Sustained
Dense MatMul	FP32	21215.19
Dense MatMul	FP16	26817.66
Dense MatMul	BF16	28534.12
Dense MatMul	FP64	69.68
Sparse MatMul	FP32	18.65
Sparse MatMul	FP64	13.13

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