Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Nemotron

Benchmarks

Task NameDataset NameSOTA ResultTrend
Binary safety classificationNemotron Response
F1 Score92
13
Binary safety classificationNemotron Query
F1 Score92
13
English to Slovene translationNemotron-Chat
COMET Score0.6975
8
Language ModelingNemotron-3-Nano (val)
Validation Loss1.4122
6
Audio Safety Guardrail AccuracyNemotron Content Safety Audio
Accuracy91.3
6
Language ModelingNemotron
Perplexity14.9
3
Latency profilingNemotron-H-8B
TPOT6.8
3
Model CompressionNemotron-8B
Model Size (GB)2.9
3
Energy consumption rankingNemotron 9B workload V2
Pairwise Accuracy96.1
2
Tool-Risk PredictionNemotron core (held-out test)
Tool-Risk Accuracy (tool rows)90.3
2
Tool-Need PredictionNemotron held-out core (test)
Tool-Need Accuracy75.3
2
Model RankingNemotron F1 (test)
Kendall Tau0.707
2
Showing 12 of 12 rows