Nemotron 3.5 Bolt 74B

2-bit GGUF + lilicorr drafter on llama.cpp. Street print on one Spark.

Params

74B

Fastest decode

80 tok/s

DGX Spark · 2-bit GGUF + lilicorr drafter · llama.cpp

Refusal

Official

Setups

1

Best reported pair · not live serving speed ↗

Run this model

Runs on

Quoted one-stream decode on real boxes.

View all

DGX Spark2-bit80 tok/s

Get it running

1. DownloadPuts the weights on this computer.
huggingface-cli download nvidia/Nemotron-3.5-Bolt
2. Start itUse this only if the download is GGUF.
llama-server -hf nvidia/Nemotron-3.5-Bolt
Open the weightsHugging Face ↗

More from NVIDIA