317 likesNew · 2d old
Tencent releases Hy4-Flash: 220B multimodal model with 1M token context
What is Hy4-flash?
Hy4-Flash is a multimodal language model that accepts both text and images and generates text output. It uses only 12 billion active parameters per token while supporting up to a million tokens of context.
Who it is for
Developers building long-form chat assistants or image-aware bots will appreciate its cheap, high-capacity inference.
The story so far
Early Signal first spotted tencent/Hy4-flash on November 19, 2026, with 317 likes. On its busiest day it gained 158 likes. The project is 2 days old. Early Signal files it under Inference, Image.