Manufacturer

NVIDIA

Nemotron-Modellfamilie von NVIDIA.

Website ↗ 🤗 Hugging Face Filter in leaderboard

12Models

567Benchmarks

567Performance

0Harness

Best token generation

2.182,7 tok/s

Nemotron-3-Nano-4B

Best prefill

44.171,1 tok/s

Nemotron-3-Nano-4B

∅ Token generation

244,5 tok/s

Average of 567 runs

∅ Prefill

3.504,1 tok/s

Average of 567 runs

Best TTFT

155 ms

Nemotron-3.5-Lightning-30B-A3B

Best harness rate

no harness run yet

🏆

Nemotron-3-Nano-4B · fastest model at 2.182,7 tok/s token generation
Best prefill of this manufacturer: 44.171,1 tok/s (Nemotron-3-Nano-4B)

Top models by token generation

Best measured generation throughput per model (tok/s), published performance benchmarks.

Nemotron-3-Nano-4B

2.182,7 tok/s

Nemotron-3-Nano-Omni-30B-A3B-Reasoning

1.791,5 tok/s

Nemotron-3-Nano-30B-A3B

1.773,5 tok/s

Nemotron-Cascade-2-30B-A3B

1.768,1 tok/s

Nemotron-3-Super-120B-A12B

600,9 tok/s

OpenReasoning-Nemotron-32B

481,0 tok/s

Llama-3.3-Nemotron-Super-49B-v1.5

364,1 tok/s

Nemotron-3.5-Lightning-30B-A3B

290,3 tok/s

Models

12 assigned

All benchmarks →

Llama-3.1-Nemotron-Ultra-253B-v1253B53 runs

4,3tok/s Ø

Min 0,1Max 62,1

Llama-3.3-Nemotron-Super-49B-v1.549B45 runs

77,6tok/s Ø

Min 2,5Max 364,1

Nemotron-3-Nano-30B-A3B30B83 runs

359,4tok/s Ø

Min 0,3Max 1.773,5

Nemotron-3-Nano-4B4B64 runs

552,8tok/s Ø

Min 0,3Max 2.182,7

Nemotron-3-Nano-Omni-30B-A3B-Reasoning30B61 runs

407,2tok/s Ø

Min 31,4Max 1.791,5

Nemotron-3-Super-120B-A12B120B56 runs

85,5tok/s Ø

Min 1,9Max 600,9

Nemotron-3-Super-120B-A12B-Q8120B0 runs

No performance data yet

Nemotron-3.5-Lightning-30B-A3B30B27 runs

186,2tok/s Ø

Min 50,7Max 290,3

Nemotron-Cascade-2-30B-A3B30B74 runs

365,5tok/s Ø

Min 0,3Max 1.768,1

Nemotron-H-47B-Reasoning-128K47B59 runs

45,9tok/s Ø

Min 3,2Max 203,5

Nemotron-Labs-3-Puzzle-75B-A9B75B0 runs

No performance data yet

OpenReasoning-Nemotron-32B32B45 runs

117,4tok/s Ø

Min 1,1Max 481,0