Can I run MiMo-V2-Flash on a Apple M5 Pro?
Not at these settings. No indexed quantization of MiMo-V2-Flash fits Apple M5 Pro at any context we compute, with q8_0 KV. The smallest shipped quantization is 58.20 GiB in weights alone, against 33.48 GiB usable. CPU offload can still run it, slowly.
Every quantization at every context
| Quant | Weights● | 4K◐ | 8K◐ | 16K◐ | 32K◐ | 64K◐ | 128K◐ |
|---|---|---|---|---|---|---|---|
| BF16 | 575.24 GiB | 576.1 | 576.3 | 576.8 | 577.8 | 579.8 | 583.8 |
| Q8_0 | 305.69 GiB | 306.5 | 306.8 | 307.3 | 308.3 | 310.3 | 314.3 |
| Q6_K | 236.11 GiB | 237.0 | 237.2 | 237.7 | 238.7 | 240.7 | 244.7 |
| Q5_K_M | 204.17 GiB | 205.0 | 205.3 | 205.8 | 206.8 | 208.7 | 212.7 |
| Q5_K_S | 198.04 GiB | 198.9 | 199.1 | 199.6 | 200.6 | 202.6 | 206.6 |
| Q4_1 | 180.12 GiB | 181.0 | 181.2 | 181.7 | 182.7 | 184.7 | 188.7 |
| I1-Q4_1 | 180.01 GiB | 180.9 | 181.1 | 181.6 | 182.6 | 184.6 | 188.6 |
| Q4_K_L | 174.72 GiB | 175.6 | 175.8 | 176.3 | 177.3 | 179.3 | 183.3 |
| Q4_K_M | 174.28 GiB | 175.1 | 175.4 | 175.9 | 176.9 | 178.9 | 182.9 |
| I1-Q4_K_M | 173.97 GiB | 174.8 | 175.1 | 175.6 | 176.6 | 178.6 | 182.5 |
| Q4_K_S | 168.15 GiB | 169.0 | 169.2 | 169.7 | 170.7 | 172.7 | 176.7 |
| Q4_0 | 165.08 GiB | 165.9 | 166.2 | 166.7 | 167.7 | 169.7 | 173.6 |
| I1-Q4_K_S | 163.33 GiB | 164.2 | 164.4 | 164.9 | 165.9 | 167.9 | 171.9 |
| I1-Q4_0 | 162.70 GiB | 163.6 | 163.8 | 164.3 | 165.3 | 167.3 | 171.3 |
| IQ4_NL | 162.41 GiB | 163.3 | 163.5 | 164.0 | 165.0 | 167.0 | 171.0 |
| IQ4_XS | 153.50 GiB | 154.4 | 154.6 | 155.1 | 156.1 | 158.1 | 162.1 |
| I1-IQ4_XS | 153.13 GiB | 154.0 | 154.2 | 154.7 | 155.7 | 157.7 | 161.7 |
| I1-Q3_K_L | 148.65 GiB | 149.5 | 149.7 | 150.2 | 151.2 | 153.2 | 157.2 |
| Q3_K_M | 137.19 GiB | 138.0 | 138.3 | 138.8 | 139.8 | 141.8 | 145.8 |
| I1-Q3_K_M | 137.19 GiB | 138.0 | 138.3 | 138.8 | 139.8 | 141.8 | 145.8 |
| Q3_K_L | 136.05 GiB | 136.9 | 137.1 | 137.6 | 138.6 | 140.6 | 144.6 |
| IQ3_M | 130.65 GiB | 131.5 | 131.7 | 132.2 | 133.2 | 135.2 | 139.2 |
| I1-IQ3_M | 125.52 GiB | 126.4 | 126.6 | 127.1 | 128.1 | 130.1 | 134.1 |
| Q3_K_S | 124.51 GiB | 125.4 | 125.6 | 126.1 | 127.1 | 129.1 | 133.1 |
| I1-IQ3_S | 123.99 GiB | 124.8 | 125.1 | 125.6 | 126.6 | 128.6 | 132.6 |
| I1-Q3_K_S | 123.96 GiB | 124.8 | 125.1 | 125.6 | 126.6 | 128.5 | 132.5 |
| UD-IQ3_XXS | 118.70 GiB | 119.5 | 119.8 | 120.3 | 121.3 | 123.3 | 127.3 |
| IQ3_XS | 117.68 GiB | 118.5 | 118.8 | 119.3 | 120.3 | 122.3 | 126.2 |
| I1-IQ3_XS | 117.12 GiB | 118.0 | 118.2 | 118.7 | 119.7 | 121.7 | 125.7 |
| IQ3_XXS | 113.14 GiB | 114.0 | 114.2 | 114.7 | 115.7 | 117.7 | 121.7 |
| I1-IQ3_XXS | 110.41 GiB | 111.3 | 111.5 | 112.0 | 113.0 | 115.0 | 119.0 |
| Q2_K_L | 104.70 GiB | 105.6 | 105.8 | 106.3 | 107.3 | 109.3 | 113.3 |
| Q2_K | 104.57 GiB | 105.4 | 105.7 | 106.2 | 107.2 | 109.1 | 113.1 |
| I1-Q2_K | 104.57 GiB | 105.4 | 105.7 | 106.2 | 107.2 | 109.1 | 113.1 |
| UD-IQ2_M | 99.33 GiB | 100.2 | 100.4 | 100.9 | 101.9 | 103.9 | 107.9 |
| I1-Q2_K_S | 97.22 GiB | 98.1 | 98.3 | 98.8 | 99.8 | 101.8 | 105.8 |
| UD-IQ2_XXS | 94.58 GiB | 95.4 | 95.7 | 96.2 | 97.2 | 99.2 | 103.1 |
| I1-IQ2_M | 93.87 GiB | 94.7 | 95.0 | 95.5 | 96.5 | 98.5 | 102.4 |
| IQ2_M | 90.09 GiB | 90.9 | 91.2 | 91.7 | 92.7 | 94.7 | 98.7 |
| UD-IQ1_M | 87.36 GiB | 88.2 | 88.5 | 89.0 | 89.9 | 91.9 | 95.9 |
| I1-IQ2_S | 85.29 GiB | 86.1 | 86.4 | 86.9 | 87.9 | 89.9 | 93.9 |
| I1-IQ2_XS | 84.00 GiB | 84.8 | 85.1 | 85.6 | 86.6 | 88.6 | 92.6 |
| UD-IQ1_S | 81.23 GiB | 82.1 | 82.3 | 82.8 | 83.8 | 85.8 | 89.8 |
| IQ2_S | 79.34 GiB | 80.2 | 80.4 | 80.9 | 81.9 | 83.9 | 87.9 |
| IQ2_XS | 79.09 GiB | 79.9 | 80.2 | 80.7 | 81.7 | 83.7 | 87.7 |
| I1-IQ2_XXS | 75.37 GiB | 76.2 | 76.5 | 77.0 | 78.0 | 80.0 | 83.9 |
| UD-TQ1_0 | 72.36 GiB | 73.2 | 73.5 | 74.0 | 74.9 | 76.9 | 80.9 |
| IQ2_XXS | 68.47 GiB | 69.3 | 69.6 | 70.1 | 71.1 | 73.1 | 77.0 |
| I1-IQ1_M | 64.64 GiB | 65.5 | 65.7 | 66.2 | 67.2 | 69.2 | 73.2 |
| IQ1_M | 61.31 GiB | 62.2 | 62.4 | 62.9 | 63.9 | 65.9 | 69.9 |
| IQ1_S | 58.91 GiB | 59.8 | 60.0 | 60.5 | 61.5 | 63.5 | 67.5 |
| I1-IQ1_S | 58.20 GiB | 59.1 | 59.3 | 59.8 | 60.8 | 62.8 | 66.8 |
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 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.