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 q8_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 | 576.4 | 576.7 | 577.2 | 578.2 | 580.2 | 584.2 |
| Q8_0 | 305.69 GiB | 306.9 | 307.1 | 307.6 | 308.6 | 310.6 | 314.6 |
| Q6_K | 236.11 GiB | 237.3 | 237.6 | 238.1 | 239.1 | 241.0 | 245.0 |
| Q5_K_M | 204.17 GiB | 205.4 | 205.6 | 206.1 | 207.1 | 209.1 | 213.1 |
| Q5_K_S | 198.04 GiB | 199.2 | 199.5 | 200.0 | 201.0 | 203.0 | 207.0 |
| Q4_1 | 180.12 GiB | 181.3 | 181.6 | 182.1 | 183.1 | 185.1 | 189.0 |
| I1-Q4_1 | 180.01 GiB | 181.2 | 181.5 | 182.0 | 182.9 | 184.9 | 188.9 |
| Q4_K_L | 174.72 GiB | 175.9 | 176.2 | 176.7 | 177.7 | 179.6 | 183.6 |
| Q4_K_M | 174.28 GiB | 175.5 | 175.7 | 176.2 | 177.2 | 179.2 | 183.2 |
| I1-Q4_K_M | 173.97 GiB | 175.2 | 175.4 | 175.9 | 176.9 | 178.9 | 182.9 |
| Q4_K_S | 168.15 GiB | 169.3 | 169.6 | 170.1 | 171.1 | 173.1 | 177.1 |
| Q4_0 | 165.08 GiB | 166.3 | 166.5 | 167.0 | 168.0 | 170.0 | 174.0 |
| I1-Q4_K_S | 163.33 GiB | 164.5 | 164.8 | 165.3 | 166.3 | 168.3 | 172.3 |
| I1-Q4_0 | 162.70 GiB | 163.9 | 164.1 | 164.6 | 165.6 | 167.6 | 171.6 |
| IQ4_NL | 162.41 GiB | 163.6 | 163.9 | 164.4 | 165.4 | 167.3 | 171.3 |
| IQ4_XS | 153.50 GiB | 154.7 | 155.0 | 155.4 | 156.4 | 158.4 | 162.4 |
| I1-IQ4_XS | 153.13 GiB | 154.3 | 154.6 | 155.1 | 156.1 | 158.1 | 162.1 |
| I1-Q3_K_L | 148.65 GiB | 149.8 | 150.1 | 150.6 | 151.6 | 153.6 | 157.6 |
| Q3_K_M | 137.19 GiB | 138.4 | 138.6 | 139.1 | 140.1 | 142.1 | 146.1 |
| I1-Q3_K_M | 137.19 GiB | 138.4 | 138.6 | 139.1 | 140.1 | 142.1 | 146.1 |
| Q3_K_L | 136.05 GiB | 137.2 | 137.5 | 138.0 | 139.0 | 141.0 | 145.0 |
| IQ3_M | 130.65 GiB | 131.9 | 132.1 | 132.6 | 133.6 | 135.6 | 139.6 |
| I1-IQ3_M | 125.52 GiB | 126.7 | 127.0 | 127.5 | 128.5 | 130.5 | 134.4 |
| Q3_K_S | 124.51 GiB | 125.7 | 126.0 | 126.5 | 127.5 | 129.4 | 133.4 |
| I1-IQ3_S | 123.99 GiB | 125.2 | 125.4 | 125.9 | 126.9 | 128.9 | 132.9 |
| I1-Q3_K_S | 123.96 GiB | 125.2 | 125.4 | 125.9 | 126.9 | 128.9 | 132.9 |
| UD-IQ3_XXS | 118.70 GiB | 119.9 | 120.1 | 120.6 | 121.6 | 123.6 | 127.6 |
| IQ3_XS | 117.68 GiB | 118.9 | 119.1 | 119.6 | 120.6 | 122.6 | 126.6 |
| I1-IQ3_XS | 117.12 GiB | 118.3 | 118.6 | 119.1 | 120.1 | 122.1 | 126.0 |
| IQ3_XXS | 113.14 GiB | 114.3 | 114.6 | 115.1 | 116.1 | 118.1 | 122.1 |
| I1-IQ3_XXS | 110.41 GiB | 111.6 | 111.9 | 112.4 | 113.3 | 115.3 | 119.3 |
| Q2_K_L | 104.70 GiB | 105.9 | 106.1 | 106.6 | 107.6 | 109.6 | 113.6 |
| Q2_K | 104.57 GiB | 105.8 | 106.0 | 106.5 | 107.5 | 109.5 | 113.5 |
| I1-Q2_K | 104.57 GiB | 105.8 | 106.0 | 106.5 | 107.5 | 109.5 | 113.5 |
| UD-IQ2_M | 99.33 GiB | 100.5 | 100.8 | 101.3 | 102.3 | 104.3 | 108.2 |
| I1-Q2_K_S | 97.22 GiB | 98.4 | 98.7 | 99.2 | 100.2 | 102.2 | 106.1 |
| UD-IQ2_XXS | 94.58 GiB | 95.8 | 96.0 | 96.5 | 97.5 | 99.5 | 103.5 |
| I1-IQ2_M | 93.87 GiB | 95.1 | 95.3 | 95.8 | 96.8 | 98.8 | 102.8 |
| IQ2_M | 90.09 GiB | 91.3 | 91.5 | 92.0 | 93.0 | 95.0 | 99.0 |
| UD-IQ1_M | 87.36 GiB | 88.6 | 88.8 | 89.3 | 90.3 | 92.3 | 96.3 |
| I1-IQ2_S | 85.29 GiB | 86.5 | 86.7 | 87.2 | 88.2 | 90.2 | 94.2 |
| I1-IQ2_XS | 84.00 GiB | 85.2 | 85.4 | 85.9 | 86.9 | 88.9 | 92.9 |
| UD-IQ1_S | 81.23 GiB | 82.4 | 82.7 | 83.2 | 84.2 | 86.2 | 90.1 |
| IQ2_S | 79.34 GiB | 80.5 | 80.8 | 81.3 | 82.3 | 84.3 | 88.3 |
| IQ2_XS | 79.09 GiB | 80.3 | 80.5 | 81.0 | 82.0 | 84.0 | 88.0 |
| I1-IQ2_XXS | 75.37 GiB | 76.6 | 76.8 | 77.3 | 78.3 | 80.3 | 84.3 |
| UD-TQ1_0 | 72.36 GiB | 73.6 | 73.8 | 74.3 | 75.3 | 77.3 | 81.3 |
| IQ2_XXS | 68.47 GiB | 69.7 | 69.9 | 70.4 | 71.4 | 73.4 | 77.4 |
| I1-IQ1_M | 64.64 GiB | 65.8 | 66.1 | 66.6 | 67.6 | 69.6 | 73.6 |
| IQ1_M | 61.31 GiB | 62.5 | 62.8 | 63.3 | 64.2 | 66.2 | 70.2 |
| IQ1_S | 58.91 GiB | 60.1 | 60.4 | 60.9 | 61.9 | 63.8 | 67.8 |
| I1-IQ1_S | 58.20 GiB | 59.4 | 59.7 | 60.1 | 61.1 | 63.1 | 67.1 |
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.