Loading…
Loading…
Models / Hermes-4
2 models (14B to 70B) · ↓ 5.5K downloads a month · from $0.32/h on Quadro RTX 8000 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 |
|---|---|---|---|---|
| OriginalNousResearch/Hermes-4-14B | vLLM | 30 GB | 36.5 GB | $0.35/h 4x RTX 3060 |
| Onnx (Prince-1)Prince-1/Hermes-4-14B-Onnx | ONNX Runtime GenAI | 30 GB | 36.8 GB | $0.32/h Quadro RTX 8000 |
| Onnx (onnx-community)onnx-community/Hermes-4-14B-Onnx | ONNX Runtime GenAI | 30 GB | 36.8 GB | $0.32/h Quadro RTX 8000 |
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| OriginalNousResearch/Hermes-4-70B | vLLM | 141 GB | 162 GB | $2.07/h 4x RTX 4090 |
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").