Native Linux

NVIDIA H100 NVL

NVIDIA · GPU · Results: 26Q3, recorded 2026-08-13

692.67
GT · rank #7 of 22

training

588.50 category score
Model Precision Sustained
MobileNetV3 FP32 537.67
MobileNetV3 BF16 243.47
Flan-T5 Small FP32 898.75
Flan-T5 Small BF16 523.52
DistilGPT2 FP32 903.48
DistilGPT2 BF16 522.66
DistilBERT FP32 1081.01
DistilBERT BF16 940.17
GraphGPS (Peptides) FP32 318.48
GraphGPS (Peptides) BF16 357.80

inference

1123.94 category score
Model Precision Sustained
MobileNetV3 FP16 2187.52
MobileNetV3 INT8 692.80
Flan-T5 Small FP16 2327.68
Flan-T5 Small INT8 947.28
Flan-T5 Small FP8 1356.97
DistilGPT2 FP16 2084.27
DistilGPT2 INT8 954.30
DistilGPT2 FP8 1194.18
DistilBERT FP16 1714.39
DistilBERT INT8 720.85
DistilBERT FP8 912.88
GraphGPS (Peptides) FP16 748.98
GraphGPS (Peptides) INT8 341.34
GraphGPS (Peptides) FP8 368.62

compute

347.75 category score
Model Precision Sustained
Dense MatMul FP32 706.80
Dense MatMul FP16 696.38
Dense MatMul BF16 715.56
Dense MatMul FP64 240.90
Sparse MatMul FP32 143.02
Sparse MatMul FP64 254.53

Results: 26Q3 · 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

The Personal Edition submits your benchmark results to TechnoLynx's servers as part of participating in the public leaderboard. If that isn't acceptable for your use case, contact TechnoLynx about Press, Pro, or Enterprise licenses.

Get Started