Nemotron-3-Nano-30B-A3B GGUF size and VRAM requirements
unsloth/Nemotron-3-Nano-30B-A3B-GGUF is a very large language model with 31.58 billion parameters, built on the nemotron_h_moe architecture. It is released under the other license and has been downloaded 16,856 times.
To run unsloth/Nemotron-3-Nano-30B-A3B-GGUF locally at a 4,096-token context, its quantized versions need between 20.37 GB (Q2_K_L, lowest quality) and 62.35 GB (BF16, highest quality) of memory, weights plus KV cache and a system margin included.
For most users the best balance is Q4_1, needing about 22.19 GB. That means unsloth/Nemotron-3-Nano-30B-A3B-GGUF fits entirely in the VRAM of a 24 GB GPU or larger, running fully on the GPU.
GGUF file size and memory by quantization
Compare real GGUF weight sizes, estimated KV cache and total memory for Q4, Q5, Q8 and every quantization published in this repository.
| Quant. | Bits | Quality | Weights | KV | Total | Speed~ | Verdict |
|---|---|---|---|---|---|---|---|
| Q2_K_L | 4.58 | Good | 16.85 GB | 2.71 GB | 20.37 GB | 3.0 t/s | Offload |
| IQ2_XXS | 4.59 | Good | 16.87 GB | 2.71 GB | 20.38 GB | 3.0 t/s | Offload |
| Q3_K_S | 4.59 | Good | 16.88 GB | 2.71 GB | 20.39 GB | 3.0 t/s | Offload |
| IQ2_M | 4.59 | Good | 16.88 GB | 2.71 GB | 20.39 GB | 3.0 t/s | Offload |
| IQ3_XXS | 4.6 | Good | 16.9 GB | 2.71 GB | 20.41 GB | 3.0 t/s | Offload |
| IQ4_XS | 4.6 | Good | 16.92 GB | 2.71 GB | 20.43 GB | 3.0 t/s | Offload |
| IQ4_NL | 4.61 | Good | 16.93 GB | 2.71 GB | 20.44 GB | 3.0 t/s | Offload |
| Q4_0 | 4.61 | Good | 16.96 GB | 2.71 GB | 20.48 GB | 2.9 t/s | Offload |
| Q2_K_XL | 5.05 | Very good | 18.55 GB | 2.71 GB | 22.06 GB | 2.7 t/s | Offload |
| Q3_K_XL | 5.05 | Very good | 18.57 GB | 2.71 GB | 22.08 GB | 2.7 t/s | Offload |
| Q3_K_M | 5.07 | Very good | 18.63 GB | 2.71 GB | 22.14 GB | 2.7 t/s | Offload |
| Q4_1 | 5.08 | Very good | 18.68 GB | 2.71 GB | 22.19 GB | 2.7 t/s | Offload |
| Q4_K_S | 5.58 | Very good | 20.51 GB | 2.71 GB | 24.02 GB | — | Insufficient |
| Q4_K_XL | 5.78 | Very good | 21.27 GB | 2.71 GB | 24.78 GB | — | Insufficient |
| Q5_K_S | 6.07 | Very good | 22.31 GB | 2.71 GB | 25.82 GB | — | Insufficient |
| Q4_K_M | 6.23 | Very good | 22.89 GB | 2.71 GB | 26.4 GB | — | Insufficient |
| Q5_K_M | 6.62 | Excellent | 24.35 GB | 2.71 GB | 27.86 GB | — | Insufficient |
| Q5_K_XL | 6.97 | Excellent | 25.62 GB | 2.71 GB | 29.13 GB | — | Insufficient |
| Q6_K | 8.49 | Excellent | 31.21 GB | 2.71 GB | 34.72 GB | — | Insufficient |
| Q6_K_XL | 8.49 | Excellent | 31.21 GB | 2.71 GB | 34.72 GB | — | Insufficient |
| Q8_0 | 8.51 | Excellent | 31.28 GB | 2.71 GB | 34.79 GB | — | Insufficient |
| Q8_K_XL | 10.25 | Excellent | 37.67 GB | 2.71 GB | 41.18 GB | — | Insufficient |
| BF16 | 16.01 | Excellent | 58.84 GB | 2.71 GB | 62.35 GB | — | Insufficient |
KV cache estimated (architecture unavailable). Speed is a rough estimate bounded by memory bandwidth.
Frequently asked questions
How much VRAM do you need to run unsloth/Nemotron-3-Nano-30B-A3B-GGUF?
You need about 22.19 GB of VRAM to run unsloth/Nemotron-3-Nano-30B-A3B-GGUF entirely on the GPU using the Q4_1 quantization (at a 4,096-token context). Smaller quantizations lower the requirement at the cost of quality.
Can I run unsloth/Nemotron-3-Nano-30B-A3B-GGUF on an 8 GB GPU?
Partially. unsloth/Nemotron-3-Nano-30B-A3B-GGUF only fits on an 8 GB GPU by offloading part of it to system RAM (with Q4_1), which runs but is slower.
Can I run unsloth/Nemotron-3-Nano-30B-A3B-GGUF on a 16 GB GPU?
Partially. unsloth/Nemotron-3-Nano-30B-A3B-GGUF only fits on a 16 GB GPU by offloading part of it to system RAM (with Q8_K_XL), which runs but is slower.
Can I run unsloth/Nemotron-3-Nano-30B-A3B-GGUF on a 24 GB GPU?
Yes. With 24 GB of VRAM you can run unsloth/Nemotron-3-Nano-30B-A3B-GGUF fully on the GPU using Q4_1 (about 22.19 GB).
What is the best quantization for unsloth/Nemotron-3-Nano-30B-A3B-GGUF?
If memory allows, higher bits-per-weight means better quality. A common sweet spot is a Q4_K_M or Q5_K_M quantization, which keeps most of the quality while roughly halving the memory versus 8-bit. Pick the highest quantization that still fits in your VRAM.