Can I run r1-1776 on a Radeon RX 6500 XT?
Not at these settings. No indexed quantization of r1-1776 fits Radeon RX 6500 XT at any context we compute, with q4_0 KV. The smallest shipped quantization is 138.66 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 | 1250.09 GiB | 1251.1 | 1251.2 | 1251.4 | 1251.7 | 1252.3 | 1253.5 |
| Q8_0 | 664.30 GiB | 665.4 | 665.4 | 665.6 | 665.9 | 666.5 | 667.7 |
| Q6_K | 512.97 GiB | 514.0 | 514.1 | 514.3 | 514.6 | 515.2 | 516.4 |
| Q5_K_M | 442.75 GiB | 443.8 | 443.9 | 444.0 | 444.3 | 444.9 | 446.1 |
| Q5_K_S | 430.10 GiB | 431.2 | 431.2 | 431.4 | 431.7 | 432.3 | 433.5 |
| Q4_1 | 391.10 GiB | 392.2 | 392.2 | 392.4 | 392.7 | 393.3 | 394.5 |
| Q4_K_M | 376.65 GiB | 377.7 | 377.8 | 377.9 | 378.2 | 378.8 | 380.0 |
| Q4_K_S | 353.90 GiB | 355.0 | 355.0 | 355.2 | 355.5 | 356.1 | 357.3 |
| Q4_0 | 353.00 GiB | 354.1 | 354.1 | 354.3 | 354.6 | 355.2 | 356.4 |
| IQ4_NL | 352.10 GiB | 353.2 | 353.2 | 353.4 | 353.7 | 354.3 | 355.5 |
| IQ4_XS | 332.60 GiB | 333.7 | 333.7 | 333.9 | 334.2 | 334.8 | 336.0 |
| Q3_K_L | 323.58 GiB | 324.6 | 324.7 | 324.9 | 325.2 | 325.8 | 327.0 |
| Q3_K_M | 297.28 GiB | 298.3 | 298.4 | 298.6 | 298.9 | 299.5 | 300.7 |
| IQ3_M | 272.03 GiB | 273.1 | 273.2 | 273.3 | 273.6 | 274.2 | 275.4 |
| Q3_K_S | 269.23 GiB | 270.3 | 270.4 | 270.5 | 270.8 | 271.4 | 272.6 |
| IQ3_XS | 254.08 GiB | 255.1 | 255.2 | 255.4 | 255.7 | 256.3 | 257.5 |
| IQ3_XXS | 240.22 GiB | 241.3 | 241.3 | 241.5 | 241.8 | 242.4 | 243.6 |
| Q2_K_L | 228.11 GiB | 229.2 | 229.2 | 229.4 | 229.7 | 230.3 | 231.5 |
| Q2_K | 227.27 GiB | 228.3 | 228.4 | 228.5 | 228.9 | 229.5 | 230.7 |
| IQ2_M | 202.50 GiB | 203.6 | 203.6 | 203.8 | 204.1 | 204.7 | 205.9 |
| IQ2_S | 183.47 GiB | 184.5 | 184.6 | 184.7 | 185.0 | 185.7 | 186.9 |
| IQ2_XS | 181.69 GiB | 182.7 | 182.8 | 183.0 | 183.3 | 183.9 | 185.1 |
| IQ2_XXS | 162.45 GiB | 163.5 | 163.6 | 163.7 | 164.0 | 164.6 | 165.8 |
| IQ1_M | 138.66 GiB | 139.7 | 139.8 | 139.9 | 140.2 | 140.8 | 142.0 |
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, this model uses latent attention and allocates no V cache at all, so any formula reading num_key_value_heads overstates its cache by more than an order of magnitude.