Loading…
Loading…
Models / SmolLM2
3 models (135M to 1.7B) · ↓ 5.7M downloads a month · from $0.08/h on GTX 1660 SUPER 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.
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| OriginalHuggingFaceTB/SmolLM2-135M | vLLM | 270 MB | 2.7 GB | $0.10/h RTX 3060 |
| InstructHuggingFaceTB/SmolLM2-135M-Instruct | vLLM | 270 MB | 2.7 GB | $0.10/h RTX 3060 |
| SmolVLM 256M InstructHuggingFaceTB/SmolVLM-256M-Instruct | vLLM | 510 MB | 3 GB | $0.10/h RTX 3060 |
| SmolDocling 256M preview (docling-project)docling-project/SmolDocling-256M-preview | vLLM | 510 MB | 3 GB | $0.10/h RTX 3060 |
| Instruct GPTQ (lilmeaty)lilmeaty/SmolLM2-135M-Instruct-GPTQ | vLLM | 90 MB | 2.5 GB | $0.10/h RTX 3060 |
| Instruct (Prince-1)Prince-1/SmolLM2-135M-Instruct | ONNX Runtime GenAI | 270 MB | 2 GB | $0.08/h GTX 1660 SUPER |
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| OriginalHuggingFaceTB/SmolLM2-360M | vLLM | 720 MB | 3.4 GB | $0.10/h RTX 3060 |
| SmolVLM2 500M Video InstructHuggingFaceTB/SmolVLM2-500M-Video-Instruct | vLLM | 1.0 GB | 3.7 GB | $0.10/h RTX 3060 |
| InstructHuggingFaceTB/SmolLM2-360M-Instruct | vLLM | 720 MB | 3.4 GB | $0.10/h RTX 3060 |
| SmolVLM 500M InstructHuggingFaceTB/SmolVLM-500M-Instruct | vLLM | 1.0 GB | 3.7 GB | $0.10/h RTX 3060 |
| Instruct ort genai int4 cpu (gianlucamazza)gianlucamazza/SmolLM2-360M-Instruct-ort-genai-int4-cpu | ONNX Runtime GenAI | 420 MB | 2.3 GB | $0.08/h GTX 1660 SUPER |
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| OriginalHuggingFaceTB/SmolLM2-1.7B | vLLM | 3.4 GB | 7.8 GB | $0.10/h RTX 3060 |
| InstructHuggingFaceTB/SmolLM2-1.7B-Instruct | vLLM | 3.4 GB | 7.8 GB | $0.10/h RTX 3060 |
| SmolVLM InstructHuggingFaceTB/SmolVLM-Instruct | vLLM | 4.5 GB | 9 GB | $0.10/h RTX 3060 |
| ONNX (haydso)haydso/SmolLM2-1.7B-ONNX | ONNX Runtime GenAI | 6.8 GB | 11 GB | $0.09/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").