training

1065.97 category score

Model	Precision	Sustained
MobileNetV3	FP32	641.26
MobileNetV3	BF16	251.52
Flan-T5 Small	FP32	2202.60
Flan-T5 Small	BF16	1085.59
DistilGPT2	FP32	1939.46
DistilGPT2	BF16	960.73
DistilBERT	FP32	2400.48
DistilBERT	BF16	2029.16
GraphGPS (Peptides)	FP32	599.94
GraphGPS (Peptides)	BF16	567.88

inference

2203.38 category score

Model	Precision	Sustained
MobileNetV3	FP16	4642.77
MobileNetV3	INT8	1590.18
Flan-T5 Small	FP16	6389.69
Flan-T5 Small	INT8	2364.50
Flan-T5 Small	FP8	2357.12
DistilGPT2	FP16	1860.53
DistilGPT2	INT8	870.08
DistilGPT2	FP8	887.76
DistilBERT	FP16	4010.21
DistilBERT	INT8	1765.34
DistilBERT	FP8	1826.30
GraphGPS (Peptides)	FP16	2274.49
GraphGPS (Peptides)	INT8	1039.57
GraphGPS (Peptides)	FP8	1102.43

compute

585.14 category score

Model	Precision	Sustained
Dense MatMul	FP32	16657.58
Dense MatMul	FP16	17441.29
Dense MatMul	BF16	17966.40
Dense MatMul	FP64	2172.30
Sparse MatMul	FP32	18.13
Sparse MatMul	FP64	32.38

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