Model comparison

DeepSeek-V4-Flash vs NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16

These two publish different quantization sets; the table below has the exact sizes.

From the file· summed bytes, KV per layer

Side by side

DeepSeek-V4-FlashNVIDIA-Nemotron-3-Ultra-550B-A55B-BF16
Parameters291B561B
Architecturedeepseek4nemotron_h_moe
Layers43
Native context1,048,576262,144
Mixture of expertsyes, 256 expertsyes, 512 experts
Quantizations published1215
Smallest quantization76.87 GiB175.12 GiB
Q4_K_M
Licencemitother

KV cache by context

the term that decides long-context viability
ContextDeepSeek-V4-FlashNVIDIA-Nemotron-3-Ultra-550B-A55B-BF16Ratio
4,0960.06 GiB
8,1920.06 GiB
16,3840.06 GiB
32,7680.06 GiB
65,5360.06 GiB
131,0720.06 GiB