MiniMax was founded in December 2021 by Yan Junjie, a former head of AI at Megvii (the Chinese computer vision company behind Face++), along with a team of researchers drawn from top Chinese tech firms. The timing was notable — launching an AI company between the release of GPT-3 and ChatGPT meant betting on foundation models before the hype cycle truly exploded. Yan had a clear thesis: build multimodal foundation models from scratch and wrap them in consumer-facing products that could achieve massive scale in the Chinese market. Unlike many Chinese AI startups that focused on enterprise sales first, MiniMax went after consumers from day one with an ambition that drew comparisons to a young ByteDance.
MiniMax's most visible product is Hailuo AI (literally "sea snail"), a consumer platform that launched as a chatbot companion app and evolved into something far more interesting. Hailuo's video generation capabilities, released in late 2024, stunned the AI community with their quality and coherence. The Hailuo AI video generator quickly became one of the most popular free video generation tools globally, producing clips that rivaled Runway Gen-3 and Kling in quality while being freely accessible. This was a deliberate strategy: use a free consumer product to demonstrate model capabilities, attract users, collect data, and build brand recognition internationally. By early 2025, Hailuo had become synonymous with accessible, high-quality AI video generation, and MiniMax was fielding interest from Hollywood studios and advertising agencies looking for cost-effective alternatives to traditional VFX.
What makes MiniMax technically interesting is the breadth of their in-house model development. Most AI startups specialize — text or image or voice. MiniMax builds across all three modalities plus video. Their language models (the abab series) have been quietly competitive with Chinese peers, though less well-known internationally than Qwen or GLM. Their speech synthesis technology powers realistic voice cloning and text-to-speech that rivals ElevenLabs in quality. And their video generation model uses a proprietary architecture that handles both text-to-video and image-to-video with strong temporal consistency and motion coherence. The company has also released MiniMax-01, a large language model with a "lightning attention" mechanism designed for extremely long contexts, signaling serious ambitions in the LLM space beyond just media generation.
MiniMax has been one of the best-funded AI startups in China. A $600 million round in mid-2024 reportedly valued the company at $2.5 billion, with backing from Alibaba, Tencent, and several sovereign wealth funds. This dual backing from China's two largest tech conglomerates is unusual — Alibaba and Tencent rarely coinvest — and signals the perceived strategic importance of MiniMax's multimodal technology. The company has been spending aggressively on GPU compute, reportedly operating thousands of A100s acquired before US export restrictions fully tightened, and supplementing with domestic alternatives. Like all Chinese AI labs, MiniMax faces the chip squeeze as a fundamental constraint, but their focus on efficient architectures and consumer-scale deployment rather than raw frontier model size has made them less dependent on the absolute latest hardware.
MiniMax occupies a peculiar position in the Chinese AI landscape: too consumer-focused to be grouped with enterprise-oriented companies like Zhipu and Baichuan, too technically ambitious to be dismissed as just another app company. Their real competition is increasingly global — Runway, Luma, and Kling in video; ElevenLabs in voice; and the major LLM providers in text. The international expansion through Hailuo AI gives them a distribution channel that most Chinese AI companies lack, since WeChat and Douyin-based distribution does not work outside China. Whether MiniMax can sustain its multi-front approach or will need to focus remains an open question, but their ability to ship high-quality products across text, voice, and video from a relatively small team has made them one of the most interesting companies to watch in global AI.