Young Tries AI

Everything on Pyramid-JiT

1 mention, newest first.

  1. Linum releases an experimental text-to-image model trained for about $21K

    Linum, a 2-person lab training generative video models from scratch, released a text-to-image model with Apache 2.0 weights. It is a pixel-space model with no VAE, using 22 DiT blocks and a frozen Qwen3.5-4B text encoder. The run is described as a proof of concept on the way to the lab's next video model.

    Usable today