Can I run North-Mini-Code-1.0 on a Radeon RX 6500 XT?
Not at these settings. No indexed quantization of North-Mini-Code-1.0 fits Radeon RX 6500 XT at any context we compute, with q8_0 KV. The smallest shipped quantization is 7.93 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 | 56.81 GiB | 57.9 | 58.0 | 58.1 | 58.3 | 58.7 | 59.6 |
| Q8_0 | 30.21 GiB | 31.3 | 31.4 | 31.5 | 31.7 | 32.1 | 33.0 |
| Q6_K_L | 24.71 GiB | 25.8 | 25.9 | 26.0 | 26.2 | 26.6 | 27.5 |
| Q6_K | 24.59 GiB | 25.7 | 25.7 | 25.8 | 26.1 | 26.5 | 27.4 |
| UD-Q6_K | 23.76 GiB | 24.8 | 24.9 | 25.0 | 25.2 | 25.7 | 26.5 |
| UD-Q5_K_M | 21.37 GiB | 22.5 | 22.5 | 22.6 | 22.8 | 23.3 | 24.1 |
| Q5_K_L | 20.47 GiB | 21.5 | 21.6 | 21.7 | 21.9 | 22.4 | 23.2 |
| Q5_K_M | 20.34 GiB | 21.4 | 21.5 | 21.6 | 21.8 | 22.3 | 23.1 |
| UD-Q5_K_S | 20.13 GiB | 21.2 | 21.3 | 21.4 | 21.6 | 22.0 | 22.9 |
| Q5_K_S | 19.70 GiB | 20.8 | 20.8 | 21.0 | 21.2 | 21.6 | 22.5 |
| Q4_1 | 17.97 GiB | 19.1 | 19.1 | 19.2 | 19.4 | 19.9 | 20.7 |
| UD-Q4_K_M | 17.88 GiB | 19.0 | 19.0 | 19.1 | 19.4 | 19.8 | 20.7 |
| Q4_K_L | 17.58 GiB | 18.7 | 18.7 | 18.8 | 19.1 | 19.5 | 20.3 |
| Q4_K_M | 17.46 GiB | 18.5 | 18.6 | 18.7 | 18.9 | 19.4 | 20.2 |
| Q4_K_S | 16.81 GiB | 17.9 | 18.0 | 18.1 | 18.3 | 18.7 | 19.6 |
| UD-Q4_K_S | 16.81 GiB | 17.9 | 18.0 | 18.1 | 18.3 | 18.7 | 19.6 |
| Q4_0 | 16.32 GiB | 17.4 | 17.5 | 17.6 | 17.8 | 18.2 | 19.1 |
| IQ4_NL | 16.30 GiB | 17.4 | 17.4 | 17.6 | 17.8 | 18.2 | 19.1 |
| IQ4_XS | 15.44 GiB | 16.5 | 16.6 | 16.7 | 16.9 | 17.4 | 18.2 |
| UD-IQ4_NL | 14.47 GiB | 15.5 | 15.6 | 15.7 | 15.9 | 16.4 | 17.2 |
| UD-IQ4_XS | 14.18 GiB | 15.3 | 15.3 | 15.4 | 15.7 | 16.1 | 17.0 |
| IQ3_M | 13.84 GiB | 14.9 | 15.0 | 15.1 | 15.3 | 15.8 | 16.6 |
| Q3_K_L | 13.74 GiB | 14.8 | 14.9 | 15.0 | 15.2 | 15.7 | 16.5 |
| UD-Q3_K_M | 13.24 GiB | 14.3 | 14.4 | 14.5 | 14.7 | 15.1 | 16.0 |
| Q3_K_M | 13.23 GiB | 14.3 | 14.4 | 14.5 | 14.7 | 15.1 | 16.0 |
| IQ3_XS | 13.23 GiB | 14.3 | 14.4 | 14.5 | 14.7 | 15.1 | 16.0 |
| Q3_K_S | 12.63 GiB | 13.7 | 13.8 | 13.9 | 14.1 | 14.5 | 15.4 |
| IQ3_XXS | 12.10 GiB | 13.2 | 13.3 | 13.4 | 13.6 | 14.0 | 14.9 |
| UD-IQ3_S | 11.89 GiB | 13.0 | 13.0 | 13.1 | 13.4 | 13.8 | 14.7 |
| UD-IQ3_XXS | 10.90 GiB | 12.0 | 12.1 | 12.2 | 12.4 | 12.8 | 13.7 |
| Q2_K_L | 10.45 GiB | 11.5 | 11.6 | 11.7 | 11.9 | 12.4 | 13.2 |
| Q2_K | 10.33 GiB | 11.4 | 11.5 | 11.6 | 11.8 | 12.2 | 13.1 |
| IQ2_M | 9.82 GiB | 10.9 | 11.0 | 11.1 | 11.3 | 11.7 | 12.6 |
| UD-IQ2_M | 9.19 GiB | 10.3 | 10.3 | 10.4 | 10.7 | 11.1 | 12.0 |
| UD-IQ2_XXS | 9.11 GiB | 10.2 | 10.3 | 10.4 | 10.6 | 11.0 | 11.9 |
| IQ2_S | 8.95 GiB | 10.0 | 10.1 | 10.2 | 10.4 | 10.9 | 11.7 |
| IQ2_XS | 8.77 GiB | 9.9 | 9.9 | 10.0 | 10.2 | 10.7 | 11.5 |
| UD-IQ1_M | 8.74 GiB | 9.8 | 9.9 | 10.0 | 10.2 | 10.6 | 11.5 |
| IQ2_XXS | 7.93 GiB | 9.0 | 9.1 | 9.2 | 9.4 | 9.8 | 10.7 |
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 4,096-token window rather than the full context.