unsloth
NVIDIA Nemotron 3 Nano Omni 30B A3B Reasoning
NVIDIA Nemotron 3 Nano Omni 30B A3B Reasoning is a 31.6B mixture-of-experts nemotron_h_moe model from unsloth, with about 3.0B 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.
Parameters31.6B / 3.0B active
Context1024K
Attentiongqa
Licenseother
52 layers · 2 KV heads · head dim 128 · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| UD-Q8_K_XL | 33.0 GB | Hugging Face file bytes |
| Q8_0 | 31.3 GB | Hugging Face file bytes |
| UD-Q6_K_XL | 31.3 GB | Hugging Face file bytes |
| UD-Q6_K | 31.3 GB | Hugging Face file bytes |
| UD-Q5_K_XL | 27.2 GB | Hugging Face file bytes |
| UD-Q5_K_M | 27.0 GB | Hugging Face file bytes |
| UD-Q5_K_S | 23.1 GB | Hugging Face file bytes |
| UD-Q4_K_XL | 22.3 GB | Hugging Face file bytes |
| UD-Q4_K_M | 22.2 GB | Hugging Face file bytes |
| UD-Q4_K_S | 21.5 GB | Hugging Face file bytes |
| UD-IQ4_XS | 18.2 GB | Hugging Face file bytes |
| IQ4_NL | 36.4 GB | Hugging Face file bytes |
| UD-Q3_K_XL | 18.2 GB | Hugging Face file bytes |
| Q3_K_M | 18.2 GB | Hugging Face file bytes |
| UD-IQ3_S | 17.5 GB | Hugging Face file bytes |
| UD-IQ3_XXS | 18.1 GB | Hugging Face file bytes |
| UD-Q2_K_XL | 17.2 GB | Hugging Face file bytes |
| UD-IQ2_XXS | 17.2 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Expert offload | UD-Q4_K_M | 59–80 |
| GeForce RTX 3060 12GB | Expert offload | UD-Q4_K_M | 73–100 |
| GeForce RTX 4060 Ti 16GB | Expert offload | UD-Q4_K_M | 59–80 |
| GeForce RTX 4090 | Tight fit | UD-IQ4_XS | 242–330 |
| GeForce RTX 5090 | Tight fit | UD-Q5_K_XL | 307–419 |
| Apple M4 Pro | Raise memory limit | UD-IQ3_S | 68–92 |
| Apple M5 | Won't fit | UD-IQ2_XXS | — |
| CPU only · DDR5-5600 | Runs on CPU | UD-Q5_K_S | 18–24 |
Smallest devices that fit it fully
- RTX 4000 Ada — Tight fit, UD-Q2_K_XL, 90–123 tok/s
- Radeon RX 7900 XT — Tight fit, UD-Q2_K_XL, 201–273 tok/s
- GeForce RTX 4090 — Tight fit, UD-IQ4_XS, 242–330 tok/s
- GeForce RTX 3090 Ti — Tight fit, UD-IQ4_XS, 242–330 tok/s
- GeForce RTX 3090 — Tight fit, UD-IQ4_XS, 225–306 tok/s
- NVIDIA L4 — Tight fit, UD-Q4_K_S, 69–93 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (UD-Q4_K_M).
llama.cpp
# Expert offload is approximate. llama.cpp can pin experts to CPU: # -ot "\\.ffn_.*_exps\\.=CPU" llama-server -hf unsloth/NVIDIA-Nemotron-3-Nano-Omni-30B-A3B-Reasoning-GGUF:UD-Q4_K_M -c 8192 -ngl 99
No matching name in the Ollama library snapshot.
LM Studio
lms get unsloth/NVIDIA-Nemotron-3-Nano-Omni-30B-A3B-Reasoning-GGUF # Then load quant UD-Q4_K_M