unsloth
NVIDIA Nemotron 3 Super 120B A12B
NVIDIA Nemotron 3 Super 120B A12B is a 120.7B mixture-of-experts nemotron_h_moe model from unsloth, with about 12.0B parameters active per token. Configured context length is 1,048,576 tokens. License: other. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters121B / 12.0B active
Context1024K
Attentiongqa
Licenseother
88 layers · 2 KV heads · head dim 128 · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| BF16 | 224.9 GB | Hugging Face file bytes |
| UD-Q8_K_XL | 123.4 GB | Hugging Face file bytes |
| Q8_0 | 119.6 GB | Hugging Face file bytes |
| UD-Q6_K_XL | 109.8 GB | Hugging Face file bytes |
| UD-Q6_K | 106.9 GB | Hugging Face file bytes |
| UD-Q5_K_XL | 100.2 GB | Hugging Face file bytes |
| UD-Q5_K_M | 100.0 GB | Hugging Face file bytes |
| UD-Q5_K_S | 83.6 GB | Hugging Face file bytes |
| UD-Q4_K_XL | 78.0 GB | Hugging Face file bytes |
| UD-Q4_K_M | 76.9 GB | Hugging Face file bytes |
| UD-Q4_K_S | 73.6 GB | Hugging Face file bytes |
| IQ4_NL | 60.1 GB | Hugging Face file bytes |
| UD-IQ4_XS | 60.1 GB | Hugging Face file bytes |
| UD-Q3_K_XL | 58.3 GB | Hugging Face file bytes |
| Q3_K_M | 57.5 GB | Hugging Face file bytes |
| UD-IQ3_S | 52.7 GB | Hugging Face file bytes |
| Q3_K_S | 57.5 GB | Hugging Face file bytes |
| UD-IQ3_XXS | 52.7 GB | Hugging Face file bytes |
| UD-Q2_K_XL | 50.9 GB | Hugging Face file bytes |
| UD-IQ2_XXS | 49.1 GB | Hugging Face file bytes |
| UD-IQ1_M | 49.1 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Won't fit | UD-IQ2_XXS | — |
| GeForce RTX 3060 12GB | Won't fit | UD-IQ2_XXS | — |
| GeForce RTX 4060 Ti 16GB | Won't fit | UD-IQ2_XXS | — |
| GeForce RTX 4090 | Won't fit | UD-IQ2_XXS | — |
| GeForce RTX 5090 | Partial offload | UD-IQ3_S | 18–25 |
| Apple M4 Pro | Won't fit | UD-IQ2_XXS | — |
| Apple M5 | Won't fit | UD-IQ2_XXS | — |
| CPU only · DDR5-5600 | Won't fit | UD-IQ2_XXS | — |
Smallest devices that fit it fully
- NVIDIA H100 SXM 80GB — Runs, IQ4_NL, 258–351 tok/s
- NVIDIA H100 PCIe 80GB — Runs, IQ4_NL, 154–210 tok/s
- NVIDIA A100 80GB SXM — Runs, IQ4_NL, 157–214 tok/s
- NVIDIA A100 80GB PCIe — Runs, IQ4_NL, 149–203 tok/s
- Apple M3 Ultra — Tight fit, IQ4_NL, 63–86 tok/s
- NVIDIA H200 SXM — Tight fit, UD-Q8_K_XL, 190–259 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (UD-IQ2_XXS).
llama.cpp
llama-server -hf unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF:UD-IQ2_XXS -c 2048 -ngl 0
No matching name in the Ollama library snapshot.
LM Studio
lms get unsloth/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF # Then load quant UD-IQ2_XXS