Does DeepSeek-V4-Pro fit in 16GB of VRAM?
Not at these settings. No indexed quantization of DeepSeek-V4-Pro fits 16GB card at any context we compute, with q4_0 KV. The smallest shipped quantization is 530.31 GiB in weights alone, against 14.88 GiB usable. CPU offload can still run it, slowly.
Every quantization at every context
| Quant | Weights● | 4K◐ | 8K◐ | 16K◐ | 32K◐ | 64K◐ | 128K◐ |
|---|---|---|---|---|---|---|---|
| Q8_0 | 1557.00 GiB | 1558.1 | 1558.2 | 1558.5 | 1559.0 | 1560.1 | 1562.2 |
| Q5_K_M | 1038.96 GiB | 1040.0 | 1040.1 | 1040.4 | 1041.0 | 1042.0 | 1044.2 |
| Q4_K_M | 885.58 GiB | 886.6 | 886.8 | 887.0 | 887.6 | 888.6 | 890.8 |
| Q3_K_M | 697.01 GiB | 698.1 | 698.2 | 698.5 | 699.0 | 700.1 | 702.2 |
| Q2_K | 530.31 GiB | 531.4 | 531.5 | 531.8 | 532.3 | 533.4 | 535.5 |
Figures are GiB of total memory: weights plus KV cache plus compute buffer and backend overhead. Weights and KV are near-exact; the overhead term is modeled. Hover any cell for the breakdown.
Why there are no speeds on this page
A capacity is not a card. Whether a model fits depends only on memory, so every figure above holds for any 16GB accelerator. How fast it runs depends on memory bandwidth, which varies several-fold between cards of the same capacity — so putting a tokens-per-second number here would be inventing one. Pick a specific card from hardware and the speed column appears.
Why other calculators disagree
A parameters × bits ÷ 8 estimate ignores two things that dominate at long context. First, the weights themselves are not the nominal rate — quantizations are mixtures, so the real file is consistently larger than the label implies. Second, most of this model's layers cache only a 128-token window rather than the full context.