DeepSeek-V4-Flash-0731 (all 284B parameters) running on a single DGX Spark.
โก ~17 tok/s generation, 40 tok/s prefill
๐ฆ One 80GB GGUF file, mixed IQ2_XXS/Q8 imatrix quant
๐ง 256-expert MoE, 6 routed per token
๐ MIT licensed
frontier model (kinda) on your desk ๐
๐ค