aiqbench takeAI Research

The article centers on an AI model: Matrix-Game 4.0 generates playable game video, with its technical report and open-sourced weights as the main news. The reported result is real-time 480p at 60 FPS on a single consumer GPU with under 150 ms input latency, which matters for readers tracking whether generative world models can run interactively. Those figures are company-reported, so independent benchmarks and the open weights will be the real test of the claims.

Source covering36Kr

What this isA model's reading of the article and related coverage, written for aiqbench. Not the publisher's words.

Kunlun Tech Founder Zhou Yahui Co-Authors Technical Report on Matrix-Game 4.0

36KrOctober 16, 20261 min read

On October 16, Kunlun Tech officially released Matrix-Game 4.0, a video-generation game model, yesterday, and simultaneously open-sourced the project code and model weights. The technical report was published today. Kunlun Tech founder Zhou Yahui personally took part in the research and development of the Matrix Game 4.0 game large model. According to the company, Matrix-Game 4.0 uses a technical architecture in which a game engine and a video generation model run in coordination. On a single NVIDIA RTX 5090 consumer graphics card, it can deliver real-time gameplay at 480p resolution and 60 FPS, with an end-to-end latency of less than 150 milliseconds from player input to on-screen response. Matrix-Game 4.0 is currently available on platforms including GitHub and Hugging Face.

Read the full article at 36Kr

Read at 36Kr ↗Ask about this article