Does command-a-plus-05-2026-bf16 fit in 8GB of VRAM?
Not at these settings. No indexed quantization of command-a-plus-05-2026-bf16 fits 8GB card at any context we compute, with q4_0 KV. The smallest shipped quantization is 45.87 GiB in weights alone, against 7.44 GiB usable. CPU offload can still run it, slowly.
Every quantization at every context
| Quant | Weights● | 4K◐ | 8K◐ | 16K◐ | 32K◐ | 64K◐ | 128K◐ |
|---|---|---|---|---|---|---|---|
| Q8_0 | 216.03 GiB | 217.0 | 217.0 | 217.1 | 217.2 | 217.5 | 218.1 |
| Q6_K | 176.79 GiB | 177.7 | 177.8 | 177.9 | 178.0 | 178.3 | 178.8 |
| Q5_K_M | 146.93 GiB | 147.9 | 147.9 | 148.0 | 148.1 | 148.4 | 149.0 |
| Q5_K_S | 141.69 GiB | 142.6 | 142.7 | 142.8 | 142.9 | 143.2 | 143.7 |
| Q4_1 | 129.29 GiB | 130.2 | 130.3 | 130.4 | 130.5 | 130.8 | 131.3 |
| Q4_K_L | 126.06 GiB | 127.0 | 127.1 | 127.1 | 127.3 | 127.5 | 128.1 |
| Q4_K_M | 125.81 GiB | 126.8 | 126.8 | 126.9 | 127.0 | 127.3 | 127.9 |
| Q4_K_S | 120.92 GiB | 121.9 | 121.9 | 122.0 | 122.1 | 122.4 | 123.0 |
| Q4_0 | 117.42 GiB | 118.4 | 118.4 | 118.5 | 118.6 | 118.9 | 119.5 |
| IQ4_NL | 116.79 GiB | 117.7 | 117.8 | 117.9 | 118.0 | 118.3 | 118.8 |
| IQ4_XS | 110.67 GiB | 111.6 | 111.7 | 111.7 | 111.9 | 112.2 | 112.7 |
| IQ3_M | 99.41 GiB | 100.4 | 100.4 | 100.5 | 100.6 | 100.9 | 101.5 |
| Q3_K_L | 99.27 GiB | 100.2 | 100.3 | 100.3 | 100.5 | 100.8 | 101.3 |
| Q3_K_M | 95.53 GiB | 96.5 | 96.5 | 96.6 | 96.7 | 97.0 | 97.6 |
| IQ3_XS | 95.16 GiB | 96.1 | 96.2 | 96.2 | 96.4 | 96.6 | 97.2 |
| Q3_K_S | 91.28 GiB | 92.2 | 92.3 | 92.3 | 92.5 | 92.8 | 93.3 |
| IQ3_XXS | 87.48 GiB | 88.4 | 88.5 | 88.5 | 88.7 | 89.0 | 89.5 |
| Q2_K_L | 74.80 GiB | 75.7 | 75.8 | 75.9 | 76.0 | 76.3 | 76.8 |
| Q2_K | 74.56 GiB | 75.5 | 75.6 | 75.6 | 75.8 | 76.0 | 76.6 |
| IQ2_M | 71.32 GiB | 72.3 | 72.3 | 72.4 | 72.5 | 72.8 | 73.4 |
| IQ2_S | 65.10 GiB | 66.0 | 66.1 | 66.2 | 66.3 | 66.6 | 67.2 |
| IQ2_XS | 63.98 GiB | 64.9 | 65.0 | 65.0 | 65.2 | 65.5 | 66.0 |
| IQ2_XXS | 57.98 GiB | 58.9 | 59.0 | 59.0 | 59.2 | 59.5 | 60.0 |
| IQ1_M | 50.59 GiB | 51.5 | 51.6 | 51.7 | 51.8 | 52.1 | 52.6 |
| IQ1_S | 45.87 GiB | 46.8 | 46.9 | 46.9 | 47.1 | 47.4 | 47.9 |
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 8GB 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 4,096-token window rather than the full context.