Added DeepSeek Flash Vision judge to
DeepSeek v4 Flash experimental with vision running (fast) on an m5 max computer and analyzing an image.
The implementation for Metal / CUDA / ROCm is finished, just doing the last tests today before releasing it.
DeepSeek has released the weights of V4 Flash Vision Exp
Multimodal model for analyzing images, screenshots, interfaces and text. It's closer to Claude Opus 4.8, the company and some testers say
China just dropped a 770 billion parameter AI model.
It’s called Hy4, and Tencent just open-sourced it. But the interesting part isn’t just the number. Hy4 uses a Mixture-of-Experts architecture, meaning it has 770B total parameters but activates only around 49B for each
You have reached the end of the archive
All of deepseek