Can I run Tinman-gemma4-companion-merged on a GeForce RTX 3050?
Not at these settings. No indexed quantization of Tinman-gemma4-companion-merged fits GeForce RTX 3050 at any context we compute, with q4_0 KV. The smallest shipped quantization is 4.94 GiB in weights alone, against 5.58 GiB usable. CPU offload can still run it, slowly.
Every quantization at every context
| Quant | Weights● | 4K◐ | 8K◐ | 16K◐ | 32K◐ | 64K◐ | 128K◐ |
|---|---|---|---|---|---|---|---|
| BF16 | 13.92 GiB | 14.8 | 14.8 | 14.8 | 14.9 | 15.0 | 15.2 |
| Q8_0 | 7.43 GiB | 8.3 | 8.3 | 8.3 | 8.4 | 8.5 | 8.8 |
| Q4_K_M | 4.94 GiB | 5.8 | 5.8 | 5.8 | 5.9 | 6.0 | 6.3 |
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.