DeepSeek V3.2 685B — VRAM Requirements
作者： Jakub Rusinowski · 最后更新： 2026年7月21日
How much GPU VRAM you need to run DeepSeek V3.2 DeepSeek V3.2 685B by DeepSeek locally, a 685B-parameter model. Figures are quantized weights + KV cache + framework overhead from a memory-bandwidth roofline model.
DeepSeek V3.2 685B needs about 414 GB VRAM at Q4_K_M.
VRAM by Quantization
| Quant | Bits/weight | Weights | Total VRAM |
|---|---|---|---|
| Q2_K | 2.63 | 225.2 GB | 226.0 GB |
| Q3_K_M | 3.41 | 292.0 GB | 292.8 GB |
| Q4_K_M | 4.83 | 413.6 GB | 414.4 GB |
| Q5_K_M | 5.67 | 485.5 GB | 486.3 GB |
| Q6_K | 6.56 | 561.7 GB | 562.5 GB |
| Q8_0 | 8.50 | 727.8 GB | 728.6 GB |
| F16 | 16.00 | 1370.0 GB | 1370.8 GB |
Switch quantization in the interactive calculator, or see the full DeepSeek V3.2 model page.
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/deepseek-v3-2-685b)Estimates only — actual VRAM varies with context length, batch size, runtime and KV-cache settings.