Models › Nemotron 3 Nano 4B
Nemotron 3 Nano 4B: VRAM needs and speed by GPU
We have 1 measured speed results for it.
3.97B parameters · GGUF files on Hugging Face
Measured speeds
| Hardware | Generate tok/s | Prompt tok/s | Quant | Backend | Test | Source |
|---|---|---|---|---|---|---|
| NVIDIA DGX Spark (GB10) | 52.9 | 2,762 | Q8_0 | CUDA FA on | tg32 | llama.cpp benches details |
Memory needed per quantization
| Quant | File size | Needs (estimate) | Smallest card it fits |
|---|---|---|---|
| Q8_0 | 4.2 GB | 5.1 GB | 8 GB cards |
| BF16 | 8.0 GB | 8.9 GB | 12 GB cards |
Expected speed on popular hardware
| Hardware | Best quant that fits | Expected tok/s |
|---|---|---|
| NVIDIA RTX 3060 | Q8_0 | 72.8 estimate |
| NVIDIA RTX 4060 Ti | Q8_0 | 57.8 estimate |
| NVIDIA RTX 3090 | Q8_0 | 155 estimate |
| NVIDIA RTX 4090 | Q8_0 | 172 estimate |
| NVIDIA RTX 5090 | Q8_0 | 271 estimate |
| AMD Radeon RX 7900 XTX | Q8_0 | 172 estimate |
| Apple M4 Pro (20-core GPU) (64 GB) | Q8_0 | 54.7 estimate |
| Apple M3 Max (40-core GPU) (64 GB) | Q8_0 | 79.9 estimate |
| AMD Ryzen AI Max+ 395 (Radeon 8060S) | Q8_0 | 49.1 estimate |
Too big for your card?
Run Nemotron 3 Nano 4B on a rented 24-80 GB GPU by the hour.
Rent a bigger GPU on RunPodAffiliate link: if you sign up, RigBench may earn a referral credit. It never affects rankings or numbers.