training

411.48 category score

Model	Precision	Sustained
MobileNetV3	FP32	363.86
MobileNetV3	BF16	175.17
Flan-T5 Small	FP32	657.00
Flan-T5 Small	BF16	353.74
DistilGPT2	FP32	657.14
DistilGPT2	BF16	358.79
DistilBERT	FP32	805.46
DistilBERT	BF16	690.91
GraphGPS (Peptides)	FP32	225.70
GraphGPS (Peptides)	BF16	230.04

inference

652.75 category score

Model	Precision	Sustained
MobileNetV3	FP16	1709.26
MobileNetV3	INT8	503.53
Flan-T5 Small	FP16	1645.66
Flan-T5 Small	INT8	759.30
Flan-T5 Small	FP8	0.00
DistilGPT2	FP16	1607.95
DistilGPT2	INT8	727.01
DistilGPT2	FP8	0.00
DistilBERT	FP16	1267.42
DistilBERT	INT8	613.96
DistilBERT	FP8	0.00
GraphGPS (Peptides)	FP16	548.43
GraphGPS (Peptides)	INT8	272.75
GraphGPS (Peptides)	FP8	0.00

compute

207.50 category score

Model	Precision	Sustained
Dense MatMul	FP32	3123.66
Dense MatMul	FP16	4314.12
Dense MatMul	BF16	4773.86
Dense MatMul	FP64	1324.32
Sparse MatMul	FP32	9.15
Sparse MatMul	FP64	16.30

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