How much VRAM does Nemotron 3 Nano 4B need?
Nemotron 3 Nano 4B needs 2.2 GB for its weights at 4-bit, 3.2 GB with an 8K context and 3.6 GB with 32K. At 8-bit, 4.9 GB.
- Parameters
- 4B
- Longest context
- 256K
- Released
- Mar 2026
- License
- NVIDIA Nemotron Open Model
- Strongest at
- Docs
Memory at each precision
| Precision | Weights | With 8K | With 32K | With 128K |
|---|---|---|---|---|
| 16-bit | 7.4 GB | 8.4 GB | 8.8 GB | 10.3 GB |
| 8-bit | 3.9 GB | 4.9 GB | 5.3 GB | 6.8 GB |
| 6-bit | 3.0 GB | 4.0 GB | 4.4 GB | 5.9 GB |
| 5-bit | 2.6 GB | 3.6 GB | 4.0 GB | 5.5 GB |
| 4-bit | 2.2 GB | 3.2 GB | 3.6 GB | 5.1 GB |
| 3-bit | 1.8 GB | 2.8 GB | 3.2 GB | 4.7 GB |
| 2-bit | 1.4 GB | 2.4 GB | 2.8 GB | 4.3 GB |
| 1.6-bit | 781 MB | 1.8 GB | 2.1 GB | 3.6 GB |
One conversation, 16-bit context cache, runtime overhead included.
What runs it
| Machine | Best fit | Context | Speed | First reply |
|---|---|---|---|---|
| RTX 3060 12 GB | 8-bit | 256K | 60 tok/s | 0.6 s |
| RTX 4060 Ti 16 GB | 8-bit | 256K | 49 tok/s | 0.3 s |
| RTX 3090 | 8-bit | 256K | 140 tok/s | 0.2 s |
| RTX 4090 | 8-bit | 256K | 148 tok/s | 0.1 s |
| RTX 5090 | 8-bit | 256K | 230 tok/s | 0.1 s |
| 2× RTX 3090 | 8-bit | 256K | 140 tok/s | 0.1 s |
| RTX PRO 6000 Blackwell | 8-bit | 256K | 230 tok/s | 0.1 s |
| M4 Max 128 GB | 8-bit | 256K | 70 tok/s | 1.2 s |
| M5 Max 128 GB | 8-bit | 256K | 77 tok/s | 0.3 s |
| Ryzen AI Max+ 395 128 GB | 8-bit | 256K | 37 tok/s | 0.7 s |
| DGX Spark 128 GB | 8-bit | 256K | 47 tok/s | 0.2 s |
| M3 Ultra 256 GB | 8-bit | 256K | 94 tok/s | 0.7 s |
| M5 Ultra 512 GB | 8-bit | 256K | 120 tok/s | 0.2 s |
| H100 SXM | 8-bit | 256K | 344 tok/s | 0.1 s |
| DGX Station GB300 | 8-bit | 256K | 491 tok/s | 0.1 s |
| 8× H200 NVL | 8-bit | 256K | 415 tok/s | 0.1 s |
Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.
How to run it
- Ollama
ollama run nemotron-3-nano:4b- llama.cpp
llama-server -hf unsloth/NVIDIA-Nemotron-3-Nano-4B-GGUF- vLLM
vllm serve nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16- MLX (Mac)
mlx_lm.chat --model mlx-community/NVIDIA-Nemotron-3-Nano-4B-4bit
Questions
How much VRAM does Nemotron 3 Nano 4B need?
Nemotron 3 Nano 4B needs 2.2 GB for its weights at 4-bit, 3.2 GB with an 8K context and 3.6 GB with 32K. At 8-bit, 4.9 GB.
Can Nemotron 3 Nano 4B run on an RTX 4090?
Yes. At 8-bit it holds 256K of context and writes about 148 tokens a second.