The AI Bench
NVIDIA · OpenMDW 1.1

Nemotron 3.5 ASR Streaming

NVIDIA’s low-delay streaming speech recognition for voice agents, 40 language-locales.

How much memory it needs

Memory to run each model, from its model card or the size of its files. Smaller versions trade some quality for less memory.

ModelVersionNeedsBusiness use
Nemotron 3.5 ASR Streaming 0.6B8-bit GGUF0.7 GBAllowed
FP322.5 GB

Which computers run it

The best model in this group for each machine.

ComputerBest versionFit
Laptop or PC, no graphics card, 32 GBNemotron 3.5 ASR Streaming 0.6B FP32Runs great
RTX 4060 · 8 GBNemotron 3.5 ASR Streaming 0.6B FP32Runs great
RTX 3060 · 12 GBNemotron 3.5 ASR Streaming 0.6B FP32Runs great
RTX 5060 Ti · 16 GBNemotron 3.5 ASR Streaming 0.6B FP32Runs great
RTX 4090 · 24 GBNemotron 3.5 ASR Streaming 0.6B FP32Runs great
RTX 5090 · 32 GBNemotron 3.5 ASR Streaming 0.6B FP32Runs great
Mac with M4 · 16 GBNemotron 3.5 ASR Streaming 0.6B FP32Runs great
Mac with M4 Pro · 48 GBNemotron 3.5 ASR Streaming 0.6B FP32Runs great
Mac with M5 Max · 128 GBNemotron 3.5 ASR Streaming 0.6B FP32Runs great
Ryzen AI Max+ 395 · 128 GBNemotron 3.5 ASR Streaming 0.6B FP32Runs great
DGX Spark · 128 GBNemotron 3.5 ASR Streaming 0.6B FP32Runs great
RTX PRO 6000 · 96 GBNemotron 3.5 ASR Streaming 0.6B FP32Runs great
Mac Studio with M5 Ultra · 256 GBNemotron 3.5 ASR Streaming 0.6B FP32Runs great
Mac Studio with M3 Ultra · 512 GBNemotron 3.5 ASR Streaming 0.6B FP32Runs great

What it is good and bad at

Good at

  • Live, low-delay transcription
  • Runs on CPU (8-bit GGUF)

Not good at

  • Newer, fewer tools support it

Start it

  • Nemotron 3.5 ASR Streaming 0.6B: runs in NeMo-Speech.cpp (GGUF) (also NeMo (PyTorch)). Model card