DeepSeek has released DeepSeek-V4.1-Flash, introducing an asymmetric model architecture designed to deliver higher intelligence at reduced operational cost. The model handles text and visual tasks using smaller active parameter counts during input processing and generation.
Alongside the launch, the lab announced a 50 percent discount during off-peak hours to balance demand, though specific dollar rates were not published in the announcement text.
Research