Llama 3.3 70B Instruct — VRAM Requirements
Autor: Jakub Rusinowski · Ostatnia aktualizacja: 8 grudnia 2024
How much GPU VRAM you need to run Llama 3.3 Llama 3.3 70B Instruct by Meta locally, a 70B-parameter model. Figures are quantized weights + KV cache + framework overhead, computed from the model's parameter count and published architecture — not a throughput model. See /pl/methodology.
Llama 3.3 70B Instruct needs about 46 GB VRAM at Q4_K_M.
VRAM by Quantization
| Quant | Bits/weight | Weights | Total VRAM |
|---|---|---|---|
| Q2_K | 2.63 | 23.0 GB | 26.5 GB |
| Q3_K_M | 3.41 | 29.8 GB | 33.3 GB |
| Q4_K_M | 4.83 | 42.3 GB | 45.7 GB |
| Q5_K_M | 5.67 | 49.6 GB | 53.1 GB |
| Q6_K | 6.56 | 57.4 GB | 60.9 GB |
| Q8_0 | 8.50 | 74.4 GB | 77.9 GB |
| F16 | 16.00 | 140.0 GB | 143.5 GB |
Switch quantization in the interactive calculator, or see the full Llama 3.3 model page.
As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.
Add this badge to your model card
Model creators: paste this into your Hugging Face model card README to link readers straight to this VRAM breakdown.
[](https://llmconfigurator.com/en/vram-calculator/llama-3.3-70b?utm_source=badge&utm_medium=referral&utm_campaign=readme_badge&utm_content=llama-3.3-70b)Estimates only — actual VRAM varies with context length, batch size, runtime and KV-cache settings.