DeepSeek V4 Preview Open-Sourced: 1.6T Total / 49B Active Params, 1M Context, Flash + Pro Variants

DeepSeek·medium signal

DeepSeek released V4 Preview on April 24 with two variants: V4-Pro (1.6T total / 49B active, rivaling top closed-source models) and V4-Flash (284B total / 13B active, fast and economical). Both support 1M context and dual thinking/non-thinking modes. V4 beats all current open models in math/STEM/coding. Available via API (deepseek-v4-flash, deepseek-v4-pro). Still in preview phase with no finalization timeline per Reuters. Not previously covered in this newsletter despite being the largest open-source model release of the quarter.