SmolLM3-3B
by HuggingFaceTB
3.1B params · text-generation · 906 likes · 137.5k downloads
SmolLM3-3B is a 3.1B parameter model. At Q4 quantization it requires 2GB of VRAM. It runs comfortably on GeForce RTX 4090 (426 tok/s), GeForce RTX 5090 (640 tok/s), M4 Max 128GB (156 tok/s).
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 426 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 640 tok/s | comfortable |
| M4 Max 128GB | 128GB | 156 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 78 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 78 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 782 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 1573 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 377 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 312 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 303 tok/s | comfortable |