Tencent Hunyuan has officially released the Hy4 Flash model, which supports multimodal input and a 1M context length. The model features 224B total parameters (including 64B Engram parameters) with 12B activated parameters, and its API pricing is significantly lower than that of Hy4.
