Skip to content
Trending storyDeveloping

Skywork open-sources Matrix-Game 4.0 interactive world model under Apache-2.0

2 articles2 sourcessince Oct 14Last article 1h ago ·

Overview

AISummary of 2 articles

Skywork AI releases Matrix-Game 4.0, an open-source interactive world model, under the Apache-2.0 license.

Inference code and a browser demo are on GitHub, with model weights on Hugging Face. The system keeps gameplay and physics in an executable engine, while a causal video model renders the visuals.

Skywork says the model runs in real time on one RTX 5090 at about 83 frames per second with one denoising step. GitHub documentation lists roughly 38 to 39 frames per second for the shooter scenes, and the demos output at 480p. QbitAI reports 60 FPS on the RTX 5090 with about 18.5 GiB of peak VRAM, a figure that differs from Skywork's 83 FPS claim.

Written by AI from the articles below · updated Oct 15, 2:04 AM ET

Check the sources:

Article timeline

The articles in this story. Times are ET.

Oct 15
  1. PandailyNews
    Skywork open-sources Matrix-Game 4.0, a game engine paired with a real-time video model

    AISkywork AI released Matrix-Game 4.0 on October 15 under the Apache-2.0 license, with inference code on GitHub and model weights on Hugging Face. The system keeps gameplay and physics in an executable engine, while a causal video model renders the visuals, and the company claims it runs in real time on one RTX 5090 at about 83 frames per second with one denoising step. GitHub documentation lists roughly 38 to 39 frames per second for the shooter scenes, and the demos output at 480p.

Oct 14
  1. QbitAINews
    Skywork's Matrix-Game 4.0 runs a game world model in real time on one RTX 5090

    AISkywork AI, a unit of Kunlun Tech, released Matrix-Game 4.0, which it says is the first game world model that runs in real time on a single consumer graphics card. A game engine maintains the world state and rules, while a video model renders the frames, and the system runs at 60 FPS on an RTX 5090 with about 18.5 GiB of peak VRAM. The team has open-sourced the model and code.

Heat trend

Not enough continuous observations to show a trend yet.