🧮 VRAM Calculator
Can your GPU run that model? Pick a size and quantization to estimate the memory you need — and see which hardware fits.
Estimated VRAM
50.1 GB
weights 42.0 + KV cache 3.5 + overhead
What can run it
- ○RTX 4060 (8 GB)8 GBtoo small
- ○RTX 4070 (12 GB)12 GBtoo small
- ○RTX 4080 / 4070 Ti S (16 GB)16 GBtoo small
- ○RTX 3090 / 4090 (24 GB)24 GBtoo small
- ○RTX 5090 (32 GB)32 GBtoo small
- ○Mac (32 GB unified)24 GBtoo small
- ○Mac (64 GB unified)48 GBtoo small
- ●Mac (128 GB unified)96 GBruns it
- ○L40S (48 GB)48 GBtoo small
- ●A100 / H100 (80 GB)80 GBruns it
- ●H200 (141 GB)141 GBruns it
- ●B200 (192 GB)192 GBruns it
- ●2× H100 (160 GB)160 GBruns it
- ●8× H100 (640 GB)640 GBruns it
Estimates: weights = parameters × bits ÷ 8, plus KV-cache that grows with context, plus ~10% runtime overhead. Real usage varies by runtime (llama.cpp, vLLM, MLX) and attention implementation — treat the numbers as a planning guide, not a guarantee.