MiniMax Launches H3 Video AI Model With Open Weights, Challenging ByteDance on Price

BusinessMiniMax Launches H3 Video AI Model With Open Weights, Challenging ByteDance on Price

Chinese artificial intelligence firm MiniMax has launched H3, its newest multimodal video generation model, pledging to break the dominance of closed-source rivals through open weights and aggressive pricing as competition in the sector intensifies.

The release sent MiniMax shares higher and positioned the company directly against ByteDance, which is rolling out its latest Seedance 2.5 model. H3 arrives as Chinese developers race to close the gap with Western AI leaders in the fast-growing field of generative video.

For Businesses & Founders
Strong brands don't stay invisible, Media coverage builds credibility, authority, and visibility.
Press releases, sponsored articles, and media exposure.
From $500

H3 ranked as the world’s most powerful AI model for video editing, according to benchmark platform Artificial Analysis. In other tasks the model showed a more mixed performance, trailing Google’s Gemini Omni Flash in text-to-video generation and ranking behind ByteDance’s Seedance 2.0 in certain categories.

By offering open weights, MiniMax allows developers to download and adapt the underlying model rather than accessing it solely through a controlled interface. The approach, paired with lower pricing, is designed to attract developers and undercut competitors that keep their systems fully proprietary.

The launch underscores how quickly the Chinese video AI market is evolving. Rivals have been unveiling new models in rapid succession, with ByteDance’s earlier Seedance releases fuelling a surge in China tech stocks and drawing attention to the country’s growing capabilities in multimodal generation.

MiniMax has emerged as one of the more prominent players in the segment. The company, backed by Alibaba Group Holding and investors from Abu Dhabi, has pursued a public listing as it scales its research and commercial ambitions.

Multimodal models such as H3 combine the ability to process and generate text, images and video within a single system, a capability seen as central to the next phase of consumer and enterprise AI applications. Video generation in particular has become a competitive battleground given its potential across advertising, entertainment and social media.

The push toward open weights marks a strategic divide in the industry. While some firms argue that open access accelerates adoption and innovation, others maintain that closed systems offer greater control over safety and monetisation.

MiniMax’s competitive pricing could pressure margins across the sector as developers weigh cost against performance. With benchmark results still favouring different models across different tasks, the market remains fragmented and fast-moving.

The rapid cadence of releases suggests further advances are likely in the months ahead, as Chinese firms compete for developer loyalty and seek to establish standards in a market where no single player has yet secured a decisive lead.

Check out our other content

Check out other tags:

Most Popular Articles