Model comparison
NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 vs DeepSeek-V4-Flash-0731
These two publish different quantization sets; the table below has the exact sizes.
From the file· summed bytes, KV per layer
Side by side
| NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 | DeepSeek-V4-Flash-0731 | |
|---|---|---|
| Parameters | 561B | 304B |
| Architecture | nemotron_h_moe | deepseek4 |
| Layers | — | 43 |
| Native context | 262,144 | 1,048,576 |
| Mixture of experts | yes, 512 experts | yes, 256 experts |
| Quantizations published | 15 | 11 |
| Smallest quantization | 175.12 GiB | 76.87 GiB |
| Q4_K_M | — | — |
| Licence | other | mit |
KV cache by context
the term that decides long-context viability
| Context | NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 | DeepSeek-V4-Flash-0731 | Ratio |
|---|---|---|---|
| 4,096 | — | 0.06 GiB | — |
| 8,192 | — | 0.06 GiB | — |
| 16,384 | — | 0.06 GiB | — |
| 32,768 | — | 0.06 GiB | — |
| 65,536 | — | 0.06 GiB | — |
| 131,072 | — | 0.06 GiB | — |