Can I run EXAONE-4.5-33B on a GeForce RTX 2080 Ti?
Not at these settings. No indexed quantization of EXAONE-4.5-33B fits GeForce RTX 2080 Ti at any context we compute, with q8_0 KV. The smallest shipped quantization is 11.57 GiB in weights alone, against 10.23 GiB usable. CPU offload can still run it, slowly.
Every quantization at every context
| Quant | Weights● | 4K◐ | 8K◐ | 16K◐ | 32K◐ | 64K◐ | 128K◐ |
|---|---|---|---|---|---|---|---|
| BF16 | 61.59 GiB | 63.0 | 63.2 | 63.5 | 64.0 | 65.1 | 67.2 |
| Q8_0 | 32.73 GiB | 34.2 | 34.3 | 34.6 | 35.1 | 36.2 | 38.3 |
| I1-Q6_K | 25.27 GiB | 26.7 | 26.9 | 27.1 | 27.7 | 28.7 | 30.9 |
| Q6_K | 25.27 GiB | 26.7 | 26.9 | 27.1 | 27.7 | 28.7 | 30.9 |
| I1-Q5_K_M | 21.87 GiB | 23.3 | 23.5 | 23.7 | 24.3 | 25.3 | 27.5 |
| Q5_K_M | 21.87 GiB | 23.3 | 23.5 | 23.7 | 24.3 | 25.3 | 27.5 |
| I1-Q5_K_S | 21.28 GiB | 22.7 | 22.9 | 23.2 | 23.7 | 24.7 | 26.9 |
| Q5_K_S | 21.28 GiB | 22.7 | 22.9 | 23.2 | 23.7 | 24.7 | 26.9 |
| I1-Q4_1 | 19.40 GiB | 20.8 | 21.0 | 21.3 | 21.8 | 22.9 | 25.0 |
| I1-Q4_K_M | 18.67 GiB | 20.1 | 20.3 | 20.5 | 21.1 | 22.1 | 24.3 |
| Q4_K_M | 18.67 GiB | 20.1 | 20.3 | 20.5 | 21.1 | 22.1 | 24.3 |
| I1-Q4_K_S | 17.65 GiB | 19.1 | 19.3 | 19.5 | 20.1 | 21.1 | 23.2 |
| Q4_K_S | 17.65 GiB | 19.1 | 19.3 | 19.5 | 20.1 | 21.1 | 23.2 |
| I1-Q4_0 | 17.58 GiB | 19.0 | 19.2 | 19.5 | 20.0 | 21.1 | 23.2 |
| IQ4_XS | 16.79 GiB | 18.2 | 18.4 | 18.7 | 19.2 | 20.3 | 22.4 |
| I1-IQ4_XS | 16.63 GiB | 18.1 | 18.2 | 18.5 | 19.0 | 20.1 | 22.2 |
| I1-Q3_K_L | 16.21 GiB | 17.6 | 17.8 | 18.1 | 18.6 | 19.7 | 21.8 |
| Q3_K_L | 16.21 GiB | 17.6 | 17.8 | 18.1 | 18.6 | 19.7 | 21.8 |
| I1-Q3_K_M | 14.97 GiB | 16.4 | 16.6 | 16.9 | 17.4 | 18.4 | 20.6 |
| Q3_K_M | 14.97 GiB | 16.4 | 16.6 | 16.9 | 17.4 | 18.4 | 20.6 |
| I1-IQ3_M | 13.92 GiB | 15.3 | 15.5 | 15.8 | 16.3 | 17.4 | 19.5 |
| I1-IQ3_S | 13.57 GiB | 15.0 | 15.2 | 15.4 | 16.0 | 17.0 | 19.2 |
| I1-Q3_K_S | 13.53 GiB | 15.0 | 15.1 | 15.4 | 15.9 | 17.0 | 19.1 |
| Q3_K_S | 13.53 GiB | 15.0 | 15.1 | 15.4 | 15.9 | 17.0 | 19.1 |
| I1-Q2_K | 11.57 GiB | 13.0 | 13.2 | 13.4 | 14.0 | 15.0 | 17.2 |
| Q2_K | 11.57 GiB | 13.0 | 13.2 | 13.4 | 14.0 | 15.0 | 17.2 |
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 4,096-token window rather than the full context.