training

5.40 category score

Model	Precision	Sustained
MobileNetV3	FP32	7.25
MobileNetV3	BF16	5.54
Flan-T5 Small	FP32	0.00
Flan-T5 Small	BF16	5.86
DistilGPT2	FP32	10.72
DistilGPT2	BF16	4.91
DistilBERT	FP32	12.47
DistilBERT	BF16	6.35
GraphGPS (Peptides)	FP32	4.81
GraphGPS (Peptides)	BF16	1.83

inference

8.56 category score

Model	Precision	Sustained
MobileNetV3	FP16	20.88
MobileNetV3	INT8	6.79
Flan-T5 Small	FP16	19.80
Flan-T5 Small	INT8	13.51
Flan-T5 Small	FP8	0.01
DistilGPT2	FP16	15.95
DistilGPT2	INT8	9.91
DistilGPT2	FP8	0.03
DistilBERT	FP16	14.90
DistilBERT	INT8	10.49
DistilBERT	FP8	0.02
GraphGPS (Peptides)	FP16	7.06
GraphGPS (Peptides)	INT8	5.22
GraphGPS (Peptides)	FP8	0.01

compute

7.05 category score

Model	Precision	Sustained
Dense MatMul	FP32	45.16
Dense MatMul	FP16	24.41
Dense MatMul	BF16	24.29
Dense MatMul	FP64	4.92
Sparse MatMul	FP32	3.60
Sparse MatMul	FP64	0.42

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