MiniMax H3 video model undercuts rivals by two-thirds on 2K pricing

MiniMax's H3 multimodal model generates 2K video at less than one-third the cost of mainstream rivals, an open-weight push that threatens ByteDance and Kuaishou's lead in China's fast-growing AI video market. The Shanghai startup, one of the country's "AI tigers," released the model on July 31 with support for text, image, video and audio inputs and native stereo sound.

"The launch consolidates MiniMax's competitive position in AI video generation, and its cost advantages will help expand enterprise customers and accelerate commercialization," Jefferies said in a research note, reiterating a Buy rating with a target price of HK$1,118.

H3 generates clips of up to 15 seconds in 2K resolution, edits existing content and transfers motion between videos using reference material in different formats. Jefferies estimates H3's pricing at 768p resolution is about half that of peers, while 2K pricing is roughly one-third. MiniMax said it plans to release the model's weights within days, extending the open-weight approach increasingly adopted by Chinese AI developers into a segment where many prominent models remain proprietary.

Shares of MiniMax (0100.HK) rose 13.15 percent to HK$230.0 on Friday, adding HK$26.80, as investors weighed the pricing advantage against intensifying competition. ByteDance's Seedance 2.0 and Kuaishou's Kling 3.0, both released this year, have drawn attention for generating sophisticated clips from combined text, images, audio and video.

Pricing math favors enterprise adoption

MiniMax targets advertising, e-commerce, product design and gaming, where video production costs are a direct input. At 2K resolution, generation costs less than one-third of comparable mainstream models per second, and at 768p less than half of competing 720p models, according to the company. MiniMax did not name the comparison models.

The cost gap matters because video generation is compute-intensive, and Chinese developers are racing to cut both price and dependence on imported chips. MiniMax said H3 was designed to work with several Chinese-made semiconductors, part of a broader push in China's AI sector to reduce reliance on U.S. suppliers. The company, founded in 2022, listed in Hong Kong in January, the second of the "AI tigers" to go public after Z.AI, and is developing a 2.7-trillion-parameter language model.

Open-weight distribution is the differentiator. Where ByteDance and Kuaishou keep their video models proprietary, MiniMax's plan to publish H3's weights lets developers download and customize the underlying system, a strategy that helped Chinese language models gain global traction. The trade-off is that open weights make it harder to monetize the model directly, pushing MiniMax toward enterprise services and API usage.

What the launch means for investors

MiniMax reached a US$4 billion valuation in 2025 after a US$300 million Series B extension. The H3 launch, priced at a fraction of rivals, could accelerate revenue from enterprise customers even as ByteDance and Kuaishou defend their lead. Jefferies' HK$1,118 target implies roughly 386 percent upside from Friday's close, a bet that cost leadership converts into market share in a segment where Chinese developers are racing to challenge U.S. rivals such as OpenAI and Google.

The commercial stakes are rising as Chinese developers crowd into video generation. MiniMax's cost advantage, if sustained, could pressure rivals to cut prices, compressing margins across the segment even as adoption grows.

This article is for informational purposes only and does not constitute investment advice.