Can I run G4-Moonlight-Dusk-26B-A4B on a GeForce RTX 5050?
Not at these settings. No indexed quantization of G4-Moonlight-Dusk-26B-A4B fits GeForce RTX 5050 at any context we compute, with q4_0 KV. The smallest shipped quantization is 7.95 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◐ |
|---|---|---|---|---|---|---|---|
| F16 | 48.41 GiB | 49.3 | 49.4 | 49.5 | 49.6 | 50.0 | 50.7 |
| Q8_0 | 25.75 GiB | 26.7 | 26.7 | 26.8 | 27.0 | 27.3 | 28.0 |
| I1-Q6_K | 21.65 GiB | 22.6 | 22.6 | 22.7 | 22.9 | 23.2 | 23.9 |
| Q6_K | 21.65 GiB | 22.6 | 22.6 | 22.7 | 22.9 | 23.2 | 23.9 |
| I1-Q5_K_M | 18.29 GiB | 19.2 | 19.2 | 19.3 | 19.5 | 19.9 | 20.6 |
| Q5_K_M | 18.29 GiB | 19.2 | 19.2 | 19.3 | 19.5 | 19.9 | 20.6 |
| I1-Q5_K_S | 17.22 GiB | 18.1 | 18.2 | 18.3 | 18.4 | 18.8 | 19.5 |
| Q5_K_S | 17.22 GiB | 18.1 | 18.2 | 18.3 | 18.4 | 18.8 | 19.5 |
| I1-Q4_K_M | 16.03 GiB | 16.9 | 17.0 | 17.1 | 17.3 | 17.6 | 18.3 |
| Q4_K_M | 16.03 GiB | 16.9 | 17.0 | 17.1 | 17.3 | 17.6 | 18.3 |
| I1-Q4_1 | 15.30 GiB | 16.2 | 16.3 | 16.3 | 16.5 | 16.9 | 17.6 |
| I1-Q4_K_S | 14.79 GiB | 15.7 | 15.7 | 15.8 | 16.0 | 16.4 | 17.1 |
| Q4_K_S | 14.79 GiB | 15.7 | 15.7 | 15.8 | 16.0 | 16.4 | 17.1 |
| I1-Q4_0 | 13.88 GiB | 14.8 | 14.8 | 14.9 | 15.1 | 15.5 | 16.2 |
| IQ4_XS | 13.46 GiB | 14.4 | 14.4 | 14.5 | 14.7 | 15.0 | 15.7 |
| I1-IQ4_XS | 13.33 GiB | 14.2 | 14.3 | 14.4 | 14.5 | 14.9 | 15.6 |
| I1-Q3_K_L | 13.17 GiB | 14.1 | 14.1 | 14.2 | 14.4 | 14.7 | 15.4 |
| Q3_K_L | 13.17 GiB | 14.1 | 14.1 | 14.2 | 14.4 | 14.7 | 15.4 |
| I1-Q3_K_M | 12.67 GiB | 13.6 | 13.6 | 13.7 | 13.9 | 14.2 | 14.9 |
| Q3_K_M | 12.67 GiB | 13.6 | 13.6 | 13.7 | 13.9 | 14.2 | 14.9 |
| I1-IQ3_M | 11.84 GiB | 12.8 | 12.8 | 12.9 | 13.1 | 13.4 | 14.1 |
| I1-IQ3_S | 11.68 GiB | 12.6 | 12.6 | 12.7 | 12.9 | 13.3 | 14.0 |
| I1-Q3_K_S | 11.68 GiB | 12.6 | 12.6 | 12.7 | 12.9 | 13.3 | 14.0 |
| Q3_K_S | 11.68 GiB | 12.6 | 12.6 | 12.7 | 12.9 | 13.3 | 14.0 |
| I1-IQ3_XS | 11.13 GiB | 12.0 | 12.1 | 12.2 | 12.4 | 12.7 | 13.4 |
| I1-IQ3_XXS | 10.84 GiB | 11.8 | 11.8 | 11.9 | 12.1 | 12.4 | 13.1 |
| I1-Q2_K_S | 10.12 GiB | 11.0 | 11.1 | 11.2 | 11.3 | 11.7 | 12.4 |
| I1-Q2_K | 10.08 GiB | 11.0 | 11.0 | 11.1 | 11.3 | 11.7 | 12.4 |
| Q2_K | 10.08 GiB | 11.0 | 11.0 | 11.1 | 11.3 | 11.7 | 12.4 |
| I1-IQ2_M | 9.96 GiB | 10.9 | 10.9 | 11.0 | 11.2 | 11.5 | 12.2 |
| I1-IQ2_S | 9.49 GiB | 10.4 | 10.4 | 10.5 | 10.7 | 11.1 | 11.8 |
| I1-IQ2_XS | 9.37 GiB | 10.3 | 10.3 | 10.4 | 10.6 | 10.9 | 11.6 |
| I1-IQ2_XXS | 8.89 GiB | 9.8 | 9.8 | 9.9 | 10.1 | 10.5 | 11.2 |
| I1-IQ1_M | 8.30 GiB | 9.2 | 9.3 | 9.3 | 9.5 | 9.9 | 10.6 |
| I1-IQ1_S | 7.95 GiB | 8.9 | 8.9 | 9.0 | 9.2 | 9.5 | 10.2 |
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 1,024-token window rather than the full context.