MoWorld Unveils Real-Time Interactive World Model
量子位 · wechat · 2026-07-08
MoCore Technology, alongside Zhejiang University and Huawei, has released MoWorld, a real-time interactive world model. The article claims it is the first fully NPU-based real-time interactive world model, boasting inference speeds up to 50FPS and deployment costs around 30% of comparable GPU solutions. MoWorld generates future world states conditioned on an initial frame, text, and camera trajectories, supporting continuous control interactions similar to W/A/S/D. A technical report has been published, with plans to open-source the weights and code, and offer services via domestic NPU super-nodes.
Related event: MoWorld: Real-Time Interactive World Model on Domestic NPUs(2 posts)→
More from Multimodal
- OCT-Bench sets 10,076 questions to test whether multimodal models really understand retinal scans — Baochen Fu · 2026-07-21
- LTX-2.3 face-and-voice LoRA training can work on 12GB VRAM with heavy tradeoffs — __alpha_____ · 2026-07-21
- Seedance 2.0 turns one reference image into a cinematic fight scene — techhalla · 2026-07-21
- Seedance 2.0 keeps character consistency across 15+ shots with just 3 prompts — techhalla · 2026-07-21
- DecartAI’s Lucy 2.5 Realtime lands on fal with live video-to-video editing — gorkem · 2026-07-21
- Google Gemini’s Omni text-to-video output is getting better, user says — michaelrabone · 2026-07-21