What struck me the most was how badly DeepSeek itself has been mutilated.
Why have all these Harness upgrades directly affected the architecture of the LLM itself, when DSH was originally supposed to exist as infrastructure within the DeepSeek ecosystem? I don’t draw, I don’t
A horizontal comparison between Hy4 and Deepseek. I think Hy4 is a more interesting model.
A horizontal comparison between Hy4 and Deepseek. I think Hy4 is a more interesting model. It seems that it is very good at work such as game scene modeling, but the work is rough and many details are always lacking polish. Is there any way to train him?
Spent 3h20m building this 3D retro TV with DeepSeek v4.1 Flash, and it cost less than $1!
You can flip through the channels. Most play AI-generated videos, but one actually streams live TV! 😁 Built using DeepSeek v4.1 Flash via gmi_cloud.
You have reached the end of the archive
All of deepseek