training

132.41 category score

Model	Precision	Sustained
MobileNetV3	FP32	175.86
MobileNetV3	BF16	120.48
Flan-T5 Small	FP32	192.84
Flan-T5 Small	BF16	108.91
DistilGPT2	FP32	171.33
DistilGPT2	BF16	94.25
DistilBERT	FP32	206.31
DistilBERT	BF16	168.07
GraphGPS (Peptides)	FP32	70.49
GraphGPS (Peptides)	BF16	76.02

inference

234.64 category score

Model	Precision	Sustained
MobileNetV3	FP16	632.59
MobileNetV3	INT8	173.73
Flan-T5 Small	FP16	505.58
Flan-T5 Small	INT8	274.31
Flan-T5 Small	FP8	307.85
DistilGPT2	FP16	272.69
DistilGPT2	INT8	157.27
DistilGPT2	FP8	156.37
DistilBERT	FP16	354.83
DistilBERT	INT8	220.59
DistilBERT	FP8	226.78
GraphGPS (Peptides)	FP16	141.02
GraphGPS (Peptides)	INT8	67.30
GraphGPS (Peptides)	FP8	71.03

compute

61.42 category score

Model	Precision	Sustained
Dense MatMul	FP32	930.91
Dense MatMul	FP16	955.32
Dense MatMul	BF16	961.30
Dense MatMul	FP64	24.76
Sparse MatMul	FP32	6.04
Sparse MatMul	FP64	4.47

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