perspectiva aiqbenchProduse AI

Kunlun Intelligence's Matrix-Game 4.0 demonstrates a significant architectural shift where video generation models directly manage world states and interaction rules, decoupling traditional game engines from rendering pipelines. By achieving real-time performance on a single consumer-grade GPU with sub-150ms latency, this development highlights the maturing efficiency of large video models for interactive applications rather than just passive content generation. The simultaneous open-sourcing of the code provides a critical benchmark for evaluating the practical constraints and capabilities of current video LLMs in low-latency environments.

Sursă care acoperă36Kr

Ce este aceastaLectura unui model asupra articolului și a acoperirii conexe, scrisă pentru aiqbench. Nu sunt cuvintele publicației.

Kunlun Intelligence Unveils World's First Consumer-Grade GPU-Powered Video Large Model Game, Opens Source Simultaneously

36Kr15 octombrie 20261 min de cititArticolul și perspectiva sunt în engleză.

On October 15, Kunlun Intelligence officially launched Matrix-Game 4.0, the world's first video large model game running on a single consumer-grade GPU, and opened the source code simultaneously. Matrix-Game 4.0 establishes a new paradigm for game world models: the game engine maintains world state and interaction rules, while the video generation model renders detailed visuals in real time. It packs a real-time gaming experience at 480p resolution and 60 FPS into a single NVIDIA RTX 5090—delivering a response latency of under 150 milliseconds from pressing a key to visual reaction.

Citește articolul complet pe 36Kr

Citește pe 36Kr ↗Întreabă despre acest articol