← Back to Explorer
Pending

NVIDIA-Nemotron-3-Super-120B-A12B-FP8 (2026)

NVIDIA-Nemotron-3-Super-120B-A12B-FP8 is a large language model released by NVIDIA, released under the other license, with a claimed knowledge cutoff of 2026-03-11. Its weights, configuration, and tokenizer files have been cryptographically hashed and are queued for on-chain attestation by Weight Registry.

Description

NVIDIA Nemotron 3 Super is the mid-size tier of NVIDIA's Nemotron 3 open-model family, released March 11, 2026 at GTC, with 120B total parameters activating roughly 12B via Mixture-of-Experts on a hybrid Mamba-Transformer architecture, aimed at complex multi-agent applications.

Metadata

Type
LLM
Issuer
NVIDIA
License
other
Size
119.54 GB
Knowledge cutoff
2026-03-11
Source
https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8
Revision
744b1880a37996c5d56bf454ae164dfd74d77c4e

Hugging Face stats

62,492 downloads on Hugging Face · 280 likes · pipeline: text-generation · library: transformers

transformerssafetensorsnemotron_htext-generationnvidiapytorchnemotron-3latent-moemtpconversational

Weight Hashes

Weights file
26 shards (Merkle root) · 119.54 GB
SHA-256677f67decb08f81eb6e09667389121ed8ccd8cd4bb9a52709e2acc1787eb9f0e
SHA3-256c778c04c78a202c07c4e176b640f5fbafad2e9118f448ee9989522a4aaf01a40
BLAKE32e713c2b89a0d708a7503c8516b18536157a9fd33645e38339e496994b6678aa
Config file
config.json · 8.24 KB
SHA-256ff5d6d643b288d4149b0bf820ecb5fe87dd9bbc08b6b811241c57840e11e30e3
SHA3-256be95bb39939feee83d70ce35f05c4e19df805d6943543a693f7cc65519d6f1df
BLAKE35761beb0955f74b71b174c80884b61e13a7883261cca97fc918f9876876df5f0
Tokenizer file
tokenizer.json · 16.29 MB
SHA-256623c34567aebb18582765289fbe23d901c62704d6518d71866e0e58db892b5b7
SHA3-2562c553f216cf81b5413fef2a64fc0558163bfc108c067254f60282cedf2f75587
BLAKE329a5eae59f68dc4d0dc2a7d67c95e24291a7ee8b88ac0afaaf2d2200abf69a23

Verify it yourself

Download the file(s) from the source below, then hash them locally (sha256sum <file> on Linux/macOS, certutil -hashfile <file> SHA256 on Windows) and compare the output against the hash(es) shown above.