Will this AI model fit on your graphics card?
Enter the model size, quantization and context length. See how much memory it needs — and whether it fits on a 16, 32 or 128 GB card.
Start with a known model
Your model
Total memory needed
7.4
GB
Good news: it fits on a 16 GB card.
Does it fit?
16 GB cardFITS
weights 4.9 · context 1.0 · overhead 1.5
14.4 GB usable · 7.0 GB free
32 GB cardFITS
weights 4.9 · context 1.0 · overhead 1.5
28.8 GB usable · 21.4 GB free
128 GB cardFITS
weights 4.9 · context 1.0 · overhead 1.5
115.2 GB usable · 107.8 GB free
model weights
context (KV cache)
overhead
free space
Details for experts
| Bits per weight | 4.85 |
|---|---|
| Model weights | 4.9 GB |
| KV cache (context) | 1.0 GB |
| Overhead | 1.5 GB |
| Total | 7.4 GB |
| Fits on 16 GB (needs ≤ 14.4 GB) | yes |
| Fits on 32 GB (needs ≤ 28.8 GB) | yes |
| Fits on 128 GB (needs ≤ 115.2 GB) | yes |
Formula: weights = params × bits ÷ 8 · KV = context ÷ 1000 × GB per 1k · overhead 1.5 GB · a model fits when its total is at most 90 % of the card's memory.