Models / SmolLM
SmolLM by HuggingFaceTB
3 models (135M to 1.7B) · ↓ 189.9K downloads a month · from $0.10/h on RTX 3060 right now. Each model's versions do the same job; quantized ones (AWQ, FP8, 4-bit) need less GPU memory and run on cheaper machines.
SmolLM 135M
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalHuggingFaceTB/SmolLM-135M | vLLM | 270 MB | 2.6 GB | $0.10/h RTX 3060 |
| InstructHuggingFaceTB/SmolLM-135M-Instruct | vLLM | 270 MB | 2.6 GB | $0.10/h RTX 3060 |
SmolLM 360M
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalHuggingFaceTB/SmolLM-360M | vLLM | 720 MB | 3.1 GB | $0.10/h RTX 3060 |
| InstructHuggingFaceTB/SmolLM-360M-Instruct | vLLM | 720 MB | 3.1 GB | $0.10/h RTX 3060 |
SmolLM 1.7B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalHuggingFaceTB/SmolLM-1.7B | vLLM | 3.4 GB | 6.5 GB | $0.10/h RTX 3060 |
| InstructHuggingFaceTB/SmolLM-1.7B-Instruct | vLLM | 3.4 GB | 6.5 GB | $0.10/h RTX 3060 |
Model details come from Hugging Face and are refreshed daily; prices follow the live offers. GPU memory is GridPup's estimate: the weights, room for an 8,192-token conversation (chat models) and the runtime's own memory. Large models are split across the GPUs of one machine ("4x RTX 4090").