Can I run MiMo-V2-Flash on a Radeon RX 6500 XT?
Not at these settings. No indexed quantization of MiMo-V2-Flash fits Radeon RX 6500 XT at any context we compute, with q4_0 KV. The smallest shipped quantization is 58.20 GiB in weights alone, against 3.72 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 | 576.3 | 576.5 | 576.7 | 577.2 | 578.3 | 580.4 |
| Q8_0 | 305.69 GiB | 306.8 | 306.9 | 307.2 | 307.7 | 308.7 | 310.9 |
| Q6_K | 236.11 GiB | 237.2 | 237.3 | 237.6 | 238.1 | 239.2 | 241.3 |
| Q5_K_M | 204.17 GiB | 205.2 | 205.4 | 205.6 | 206.2 | 207.2 | 209.3 |
| Q5_K_S | 198.04 GiB | 199.1 | 199.2 | 199.5 | 200.0 | 201.1 | 203.2 |
| Q4_1 | 180.12 GiB | 181.2 | 181.3 | 181.6 | 182.1 | 183.2 | 185.3 |
| I1-Q4_1 | 180.01 GiB | 181.1 | 181.2 | 181.5 | 182.0 | 183.1 | 185.2 |
| Q4_K_L | 174.72 GiB | 175.8 | 175.9 | 176.2 | 176.7 | 177.8 | 179.9 |
| Q4_K_M | 174.28 GiB | 175.4 | 175.5 | 175.8 | 176.3 | 177.3 | 179.5 |
| I1-Q4_K_M | 173.97 GiB | 175.0 | 175.2 | 175.4 | 176.0 | 177.0 | 179.1 |
| Q4_K_S | 168.15 GiB | 169.2 | 169.4 | 169.6 | 170.2 | 171.2 | 173.3 |
| Q4_0 | 165.08 GiB | 166.2 | 166.3 | 166.6 | 167.1 | 168.1 | 170.2 |
| I1-Q4_K_S | 163.33 GiB | 164.4 | 164.5 | 164.8 | 165.3 | 166.4 | 168.5 |
| I1-Q4_0 | 162.70 GiB | 163.8 | 163.9 | 164.2 | 164.7 | 165.8 | 167.9 |
| IQ4_NL | 162.41 GiB | 163.5 | 163.6 | 163.9 | 164.4 | 165.5 | 167.6 |
| IQ4_XS | 153.50 GiB | 154.6 | 154.7 | 155.0 | 155.5 | 156.6 | 158.7 |
| I1-IQ4_XS | 153.13 GiB | 154.2 | 154.3 | 154.6 | 155.1 | 156.2 | 158.3 |
| I1-Q3_K_L | 148.65 GiB | 149.7 | 149.9 | 150.1 | 150.6 | 151.7 | 153.8 |
| Q3_K_M | 137.19 GiB | 138.3 | 138.4 | 138.7 | 139.2 | 140.2 | 142.4 |
| I1-Q3_K_M | 137.19 GiB | 138.3 | 138.4 | 138.7 | 139.2 | 140.2 | 142.4 |
| Q3_K_L | 136.05 GiB | 137.1 | 137.3 | 137.5 | 138.0 | 139.1 | 141.2 |
| IQ3_M | 130.65 GiB | 131.7 | 131.9 | 132.1 | 132.7 | 133.7 | 135.8 |
| I1-IQ3_M | 125.52 GiB | 126.6 | 126.7 | 127.0 | 127.5 | 128.6 | 130.7 |
| Q3_K_S | 124.51 GiB | 125.6 | 125.7 | 126.0 | 126.5 | 127.6 | 129.7 |
| I1-IQ3_S | 123.99 GiB | 125.1 | 125.2 | 125.5 | 126.0 | 127.0 | 129.2 |
| I1-Q3_K_S | 123.96 GiB | 125.0 | 125.2 | 125.4 | 126.0 | 127.0 | 129.1 |
| UD-IQ3_XXS | 118.70 GiB | 119.8 | 119.9 | 120.2 | 120.7 | 121.8 | 123.9 |
| IQ3_XS | 117.68 GiB | 118.8 | 118.9 | 119.2 | 119.7 | 120.7 | 122.8 |
| I1-IQ3_XS | 117.12 GiB | 118.2 | 118.3 | 118.6 | 119.1 | 120.2 | 122.3 |
| IQ3_XXS | 113.14 GiB | 114.2 | 114.4 | 114.6 | 115.1 | 116.2 | 118.3 |
| I1-IQ3_XXS | 110.41 GiB | 111.5 | 111.6 | 111.9 | 112.4 | 113.5 | 115.6 |
| Q2_K_L | 104.70 GiB | 105.8 | 105.9 | 106.2 | 106.7 | 107.8 | 109.9 |
| Q2_K | 104.57 GiB | 105.6 | 105.8 | 106.0 | 106.6 | 107.6 | 109.7 |
| I1-Q2_K | 104.57 GiB | 105.6 | 105.8 | 106.0 | 106.6 | 107.6 | 109.7 |
| UD-IQ2_M | 99.33 GiB | 100.4 | 100.5 | 100.8 | 101.3 | 102.4 | 104.5 |
| I1-Q2_K_S | 97.22 GiB | 98.3 | 98.4 | 98.7 | 99.2 | 100.3 | 102.4 |
| UD-IQ2_XXS | 94.58 GiB | 95.7 | 95.8 | 96.1 | 96.6 | 97.6 | 99.7 |
| I1-IQ2_M | 93.87 GiB | 95.0 | 95.1 | 95.3 | 95.9 | 96.9 | 99.0 |
| IQ2_M | 90.09 GiB | 91.2 | 91.3 | 91.6 | 92.1 | 93.2 | 95.3 |
| UD-IQ1_M | 87.36 GiB | 88.4 | 88.6 | 88.8 | 89.4 | 90.4 | 92.5 |
| I1-IQ2_S | 85.29 GiB | 86.4 | 86.5 | 86.8 | 87.3 | 88.3 | 90.5 |
| I1-IQ2_XS | 84.00 GiB | 85.1 | 85.2 | 85.5 | 86.0 | 87.1 | 89.2 |
| UD-IQ1_S | 81.23 GiB | 82.3 | 82.4 | 82.7 | 83.2 | 84.3 | 86.4 |
| IQ2_S | 79.34 GiB | 80.4 | 80.5 | 80.8 | 81.3 | 82.4 | 84.5 |
| IQ2_XS | 79.09 GiB | 80.2 | 80.3 | 80.6 | 81.1 | 82.1 | 84.3 |
| I1-IQ2_XXS | 75.37 GiB | 76.4 | 76.6 | 76.8 | 77.4 | 78.4 | 80.5 |
| UD-TQ1_0 | 72.36 GiB | 73.4 | 73.6 | 73.8 | 74.4 | 75.4 | 77.5 |
| IQ2_XXS | 68.47 GiB | 69.6 | 69.7 | 69.9 | 70.5 | 71.5 | 73.6 |
| I1-IQ1_M | 64.64 GiB | 65.7 | 65.9 | 66.1 | 66.6 | 67.7 | 69.8 |
| IQ1_M | 61.31 GiB | 62.4 | 62.5 | 62.8 | 63.3 | 64.4 | 66.5 |
| IQ1_S | 58.91 GiB | 60.0 | 60.1 | 60.4 | 60.9 | 62.0 | 64.1 |
| I1-IQ1_S | 58.20 GiB | 59.3 | 59.4 | 59.7 | 60.2 | 61.3 | 63.4 |
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.