Model comparison

DeepSeek-V4-Flash-0731 vs DeepSeek-V4-Flash

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-Flash-0731DeepSeek-V4-Flash
Parameters304B291B
Architecturedeepseek4deepseek4
Layers4343
Native context1,048,5761,048,576
Mixture of expertsyes, 256 expertsyes, 256 experts
Quantizations published1112
Smallest quantization76.87 GiB76.87 GiB
Q4_K_M
Licencemitmit

KV cache by context

the term that decides long-context viability
ContextDeepSeek-V4-Flash-0731DeepSeek-V4-FlashRatio
4,0960.06 GiB0.06 GiB1.00×
8,1920.06 GiB0.06 GiB1.00×
16,3840.06 GiB0.06 GiB1.00×
32,7680.06 GiB0.06 GiB1.00×
65,5360.06 GiB0.06 GiB1.00×
131,0720.06 GiB0.06 GiB1.00×