Can I run Laguna-S-2.1 on a Radeon RX 7400?
Not at these settings. No indexed quantization of Laguna-S-2.1 fits Radeon RX 7400 at any context we compute, with q8_0 KV. The smallest shipped quantization is 23.15 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◐ |
|---|---|---|---|---|---|---|---|
| BF16 | 219.05 GiB | 220.1 | 220.2 | 220.4 | 220.8 | 221.6 | 223.2 |
| F16 | 219.05 GiB | 220.1 | 220.2 | 220.4 | 220.8 | 221.6 | 223.2 |
| Q8_0 | 119.91 GiB | 121.0 | 121.1 | 121.3 | 121.7 | 122.5 | 124.1 |
| Q6_K_L | 94.83 GiB | 95.9 | 96.0 | 96.2 | 96.6 | 97.4 | 99.0 |
| UD-Q6_K | 91.19 GiB | 92.3 | 92.4 | 92.6 | 93.0 | 93.8 | 95.4 |
| Q6_K | 89.93 GiB | 91.0 | 91.1 | 91.3 | 91.7 | 92.5 | 94.1 |
| Q4_K_M | 89.44 GiB | 90.5 | 90.6 | 90.8 | 91.2 | 92.0 | 93.6 |
| UD-Q5_K_M | 81.83 GiB | 82.9 | 83.0 | 83.2 | 83.6 | 84.4 | 86.0 |
| Q5_K_L | 78.39 GiB | 79.5 | 79.6 | 79.8 | 80.2 | 81.0 | 82.6 |
| Q5_K_M | 78.22 GiB | 79.3 | 79.4 | 79.6 | 80.0 | 80.8 | 82.4 |
| UD-Q5_K_S | 76.98 GiB | 78.1 | 78.2 | 78.4 | 78.8 | 79.6 | 81.2 |
| Q5_K_S | 75.72 GiB | 76.8 | 76.9 | 77.1 | 77.5 | 78.3 | 79.9 |
| Q4_1 | 68.96 GiB | 70.1 | 70.2 | 70.4 | 70.8 | 71.5 | 73.1 |
| UD-Q4_K_M | 68.10 GiB | 69.2 | 69.3 | 69.5 | 69.9 | 70.7 | 72.3 |
| Q4_K_S | 64.36 GiB | 65.5 | 65.6 | 65.8 | 66.2 | 67.0 | 68.5 |
| UD-Q4_K_S | 63.88 GiB | 65.0 | 65.1 | 65.3 | 65.7 | 66.5 | 68.1 |
| Q4_0 | 62.44 GiB | 63.5 | 63.6 | 63.8 | 64.2 | 65.0 | 66.6 |
| IQ4_NL | 62.32 GiB | 63.4 | 63.5 | 63.7 | 64.1 | 64.9 | 66.5 |
| IQ4_XS | 58.98 GiB | 60.1 | 60.2 | 60.4 | 60.8 | 61.6 | 63.2 |
| UD-IQ4_NL | 54.71 GiB | 55.8 | 55.9 | 56.1 | 56.5 | 57.3 | 58.9 |
| UD-IQ4_XS | 53.61 GiB | 54.7 | 54.8 | 55.0 | 55.4 | 56.2 | 57.8 |
| IQ3_M | 52.60 GiB | 53.7 | 53.8 | 54.0 | 54.4 | 55.2 | 56.8 |
| Q3_K_L | 52.33 GiB | 53.4 | 53.5 | 53.7 | 54.1 | 54.9 | 56.5 |
| Q3_K_M | 50.33 GiB | 51.4 | 51.5 | 51.7 | 52.1 | 52.9 | 54.5 |
| UD-Q3_K_M | 50.31 GiB | 51.4 | 51.5 | 51.7 | 52.1 | 52.9 | 54.5 |
| IQ3_XS | 50.29 GiB | 51.4 | 51.5 | 51.7 | 52.1 | 52.9 | 54.5 |
| Q3_K_S | 48.03 GiB | 49.1 | 49.2 | 49.4 | 49.8 | 50.6 | 52.2 |
| IQ3_XXS | 46.07 GiB | 47.2 | 47.3 | 47.5 | 47.9 | 48.7 | 50.3 |
| UD-IQ3_S | 45.10 GiB | 46.2 | 46.3 | 46.5 | 46.9 | 47.7 | 49.3 |
| UD-IQ3_XXS | 41.24 GiB | 42.3 | 42.4 | 42.6 | 43.0 | 43.8 | 45.4 |
| Q2_K_L | 39.20 GiB | 40.3 | 40.4 | 40.6 | 41.0 | 41.8 | 43.4 |
| Q2_K | 38.92 GiB | 40.0 | 40.1 | 40.3 | 40.7 | 41.5 | 43.1 |
| IQ2_M | 37.19 GiB | 38.3 | 38.4 | 38.6 | 39.0 | 39.8 | 41.4 |
| UD-IQ2_M | 34.71 GiB | 35.8 | 35.9 | 36.1 | 36.5 | 37.3 | 38.9 |
| UD-IQ2_XXS | 34.64 GiB | 35.7 | 35.8 | 36.0 | 36.4 | 37.2 | 38.8 |
| IQ2_S | 33.77 GiB | 34.9 | 35.0 | 35.2 | 35.6 | 36.4 | 38.0 |
| UD-IQ1_M | 33.19 GiB | 34.3 | 34.4 | 34.6 | 35.0 | 35.8 | 37.4 |
| IQ2_XS | 33.12 GiB | 34.2 | 34.3 | 34.5 | 34.9 | 35.7 | 37.3 |
| UD-IQ1_S | 31.45 GiB | 32.5 | 32.6 | 32.8 | 33.2 | 34.0 | 35.6 |
| IQ2_XXS | 29.82 GiB | 30.9 | 31.0 | 31.2 | 31.6 | 32.4 | 34.0 |
| IQ1_M | 25.75 GiB | 26.8 | 26.9 | 27.1 | 27.5 | 28.3 | 29.9 |
| IQ1_S | 23.15 GiB | 24.2 | 24.3 | 24.5 | 24.9 | 25.7 | 27.3 |
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 512-token window rather than the full context.