training

156.52 category score

Model	Precision	Sustained
MobileNetV3	FP32	236.34
MobileNetV3	BF16	157.00
Flan-T5 Small	FP32	183.87
Flan-T5 Small	BF16	104.31
DistilGPT2	FP32	205.99
DistilGPT2	BF16	128.46
DistilBERT	FP32	247.61
DistilBERT	BF16	207.70
GraphGPS (Peptides)	FP32	83.83
GraphGPS (Peptides)	BF16	90.33

inference

228.56 category score

Model	Precision	Sustained
MobileNetV3	FP16	793.50
MobileNetV3	INT8	162.39
Flan-T5 Small	FP16	529.33
Flan-T5 Small	INT8	313.93
Flan-T5 Small	FP8	0.00
DistilGPT2	FP16	396.87
DistilGPT2	INT8	190.13
DistilGPT2	FP8	0.00
DistilBERT	FP16	438.39
DistilBERT	INT8	272.63
DistilBERT	FP8	0.00
GraphGPS (Peptides)	FP16	225.29
GraphGPS (Peptides)	INT8	75.09
GraphGPS (Peptides)	FP8	0.00

compute

72.46 category score

Model	Precision	Sustained
Dense MatMul	FP32	1121.95
Dense MatMul	FP16	1062.49
Dense MatMul	BF16	1069.76
Dense MatMul	FP64	32.62
Sparse MatMul	FP32	7.16
Sparse MatMul	FP64	5.62

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