Loading…
Loading…
Models / Llama-3.1
3 models (8B to 405B) · ↓ 55.4K downloads a month · from $0.16/h on RTX 4060 Ti 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 |
|---|---|---|---|---|
| Originalnvidia/Llama-3.1-8B-Instruct-FP8 | vLLM | 9.1 GB | 13.5 GB | $0.16/h RTX 4060 Ti |
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| Originalnvidia/Llama-3.1-70B-Instruct-FP8 | vLLM | 73 GB | 85.9 GB | $0.89/h 2x Nvidia A40 |
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| Originalnvidia/Llama-3.1-405B-Instruct-FP8 | vLLM | 410 GB | 462.5 GB | $11.77/h 8x RTX PRO 6000 |
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").