Chinese AI company MiniMax introduced a new video generation model named Hailuo 02. The latest release features native 1080p video output, advanced instruction adherence, and improved handling of extreme physical scenarios.
According to collaborators involved in early experimental testing, the model demonstrates a unique ability to render intricate physical movements—such as those seen in gymnastics—with a level of detail that is not currently matched by other available models. The company has encouraged broader exploration of its capabilities by the creative community.
Day 2/5 of #MiniMaxWeek: Introducing Hailuo 02, World-Class Quality, Record-Breaking Cost Efficiency 
– Best-in-class instruction following
– Handles extreme physics (yes, it does acrobatics 
)
– Native 1080p pic.twitter.com/OQgNt17rDF
— MiniMax (official) (@MiniMax__AI) June 17, 2025
MiniMax began development in August of the previous year by launching a demo version of its initial video generation model on a web-based platform. The prototype drew attention from global creators, prompting the development of the Hailuo Video product, which has since been used to generate more than 370 million videos.
Guided by a commitment to accessibility, MiniMax developed a new architecture, Noise-aware Compute Redistribution (NCR), to increase model efficiency without increasing costs. At similar parameter scales, this design has improved training and inference efficiency by a factor of 2.5. The increased efficiency has enabled deployment of a model with three times more parameters than its predecessor, without raising usage costs.
The model is presently positioned in second place on the Artificial Analysis leaderboard, surpassing Veo 3, and is noted for producing 1080p resolution outputs while demonstrating strong capabilities in simulating complex physical dynamics.
This expanded model capacity allows for deeper learning from data, supported by a training dataset that is four times larger and more diverse than the previous version. Combined with structural and data improvements, Hailuo 02 shows marked advancements in prompt interpretation, instruction execution, and physical simulation.
The result is a more capable and scalable model, delivering native 1080p video generation with enhanced responsiveness to complex inputs. MiniMax positions this update as a step forward in enabling broader, cost-effective access to high-quality video creation tools.
The latest improvements to the model have been fully deployed across the Hailuo Video web interface, mobile application, and open platform. Three video generation formats are currently available: 768p for 6 seconds, 768p for 10 seconds, and 1080p for 6 seconds. Due to architectural advancements, these services are offered at pricing levels that remain among the most competitive in the market, both within China and compared to global providers, including established technology firms and newer entrants.
MiniMax Emerges As AI Powerhouse With New M1 Model, Backed By Tencent And Alibaba
MiniMax is a Shanghai-based AI startup founded in December 2021, recognized as a leading “AI tiger” in China. It develops large-scale multimodal AI models—including text, voice, image, and notably video generation (like Hailuo—and generates billions of tokens and millions of video clips—all supported by major investors like Alibaba, Tencent, and IDG. The firm, backed by Tencent Holdings and Alibaba Group, is part of an elite group of domestic AI startups known as the Little Dragons, which have raised billions of dollars in venture funding over the past year.
Earlier this week, MiniMax introduced a new large language model (LLM), MiniMax-M1, which it claims is more efficient than closed-source competitors in China and outperforms DeepSeek’s latest R1-0528 model in several benchmarks. The M1 model supports a context length of one million tokens—eight times that of DeepSeek R1—and requires only about 30% of the resources DeepSeek needs under certain conditions.
The post MiniMax Unveils Hailuo 02: A Powerful Advancement In Video AI appeared first on Metaverse Post.