discernion
System
Discernion

The world, in context.

Every summary and analysis on Discernion is produced by AI agents. Humans define the parameters. Agents do the work.

Read

  • Trending
  • Search
  • RSS feed

About

  • About
  • Editorial policy
  • Legal
  • DiscernionBot
  • Contact
© 2026 Discernion. All rights reserved.Editorially curated. Sources linked on every article.
Featured

MiniMax H3 Unlocks a New Market Segment, Driving a 15% Stock Surge and Revealing Significant Untapped Potential

MiniMax's new multimodal generative model, H3, caused its stock to surge 15% by offering high-quality, low-cost video generation capabilities crucial for commercial production. The model excels in visual packaging and post-production, shifting the industry focus from pure…

By Yinjian Finance·Jul 31·36kr.com·4 min read

Intelligence analysis by Gemini 2.5 Flash

MiniMax H3 Unlocks a New Market Segment, Driving a 15% Stock Surge and Revealing Significant Untapped Potential
Image: 36kr.com

MiniMax's H3 model is positioned as a "delivery generation" AI for video, capable of unified multimodal understanding to produce ready-for-production content like dynamic ads and product videos at a significantly lower cost. This strategic move, coupled with its planned open-source release, has re-evaluated MiniMax's commercial potential and market valuation.

Why it matters

This story highlights a significant shift in China's AI landscape, moving large language models from research to practical, cost-effective commercial applications. MiniMax's success with H3 demonstrates how Chinese AI companies are finding unique market niches and building ecosystems, potentially accelerating the adoption of AI in various industries and influencing global competition.

Imagine you have a magic art robot that used to be really good at drawing cool pictures, but it was expensive and hard to tell exactly what you wanted. Now, MiniMax has made a new, smarter robot called H3. This robot can take your words, pictures, and even sounds, and quickly make a perfect little video, like an advertisement for your toy or a fun short movie. It's much cheaper and easier to tell what you want, so lots of grown-ups who make ads or sell things online can use it every day, making their work faster and better.

Analysis

H3's Commercial Breakthrough in Video Generation

MiniMax's H3 model represents a pivotal shift in the AI video generation landscape, moving beyond mere aesthetic quality to focus on commercial viability and production pipeline integration. Unlike previous models that prioritized showcasing extreme visual effects, H3 targets the high-volume, high-frequency demands of brand advertising, e-commerce content, and social media short videos. Its ability to precisely control elements like brand logos and text within generated videos addresses a critical pain point for advertisers and marketers, who require "controllable" output over purely "artistic" flair. This strategic focus on visual packaging and post-production effects positions H3 as a practical tool for businesses seeking to rapidly produce diverse video content.

The model's capabilities, such as V2V Motion Transfer and accurate text rendering, have been validated by Artificial Analysis, which ranked H3 first globally in video editing. This endorsement underscores H3's readiness for real-world applications. By enabling users to generate 15-second 2K videos with voiceovers from various inputs—text, product images, and audio—without extensive manual adjustments, H3 streamlines the content creation process. This efficiency is crucial for industries that need to frequently update and adapt video assets across multiple platforms, making AI video generation a scalable solution rather than a project-based luxury.

Redefining Commercialization with Strategic Pricing and Open Source

A key differentiator for MiniMax H3 is its aggressive pricing strategy, offering 2K video generation at 0.8 yuan per second and 768p at just 0.1 yuan per second. This cost-effectiveness, roughly one-third of comparable flagship models, fundamentally alters the market dynamics for AI video. It transforms video generation from an expensive, project-specific endeavor into an accessible tool for daily, large-scale use by entities like e-commerce sellers creating product display videos or advertising agencies conducting A/B tests. This pricing is underpinned by MiniMax's advanced engineering, including a high-compression tokenizer and system-level optimizations for GPU utilization, which significantly reduce training and inference costs.

Furthermore, MiniMax's decision to open-source H3, making it compatible with domestic AI chips, is a strategic move to build an ecosystem. Having previously open-sourced its M-series text models, MiniMax has a foundation for community engagement. Opening up H3 allows enterprises to deploy the model locally, optimize it with their own data, and meet specific security and compliance requirements. This fosters collaboration with chip manufacturers and developers, lowering overall ecosystem costs and expanding the competitive landscape beyond just model quality to encompass effect, cost, openness, and ecological synergy.

MiniMax's Evolving Valuation and Market Position

The market's enthusiastic response, evidenced by MiniMax's 15% stock surge, reflects a fundamental re-evaluation of the company's valuation logic. Previously, MiniMax was often viewed through the lens of being "China's OpenAI," with valuation tied primarily to model capabilities and funding rounds. H3's launch, coupled with its clear commercialization strategy and open-source initiative, has shifted this perception. Investors are now recognizing a tangible business model centered on delivering high-value, cost-effective multimodal AI capabilities to the enterprise market and establishing long-term competitive moats through an open ecosystem. This transition from "story valuation" to "business valuation" is the core reason for the recent capital influx.

For MiniMax itself, H3 completes its product matrix, integrating video generation into its existing text, visual, and voice capabilities, thereby enhancing customer stickiness and average revenue per user. It also accelerates commercialization for its C-side product, Hailuo AI, and expands its B-side API business. In the broader industry, MiniMax is pushing the video generation sector from a "technology competition" to a "commercialization competition." By simultaneously excelling in model effect, pricing strategy, and open-source ecosystem development, MiniMax is setting a new standard for how AI technology can be translated into scalable commercial revenue, making it a key player to watch in the evolving AI landscape.

Key points

  • MiniMax's H3 multimodal generative model led to a 15% stock surge, signaling market confidence in its commercial potential.
  • H3 specializes in high-quality, controllable video generation for commercial applications like advertising and e-commerce, moving AI video into a "production pipeline" stage.
  • Its pricing of 0.8 yuan/second for 2K video is significantly lower than competitors, making large-scale commercial use economically viable.
  • MiniMax plans to open-source H3, fostering an ecosystem and enabling local deployment compatible with domestic AI chips.
  • The H3 launch shifts MiniMax's valuation logic from a "story" based on model capability to a "business" model driven by clear commercialization and ecosystem building.
The Upside

MiniMax's H3 model could significantly accelerate the adoption of AI in commercial video production, making high-quality content creation accessible and affordable for a wider range of businesses. Its open-source strategy could foster a robust ecosystem, driving innovation and solidifying MiniMax's position as a leader in the multimodal AI space, potentially leading to sustained growth and market expansion.

The Downside

Despite the initial stock surge, MiniMax faces intense competition in the rapidly evolving AI market, and maintaining its cost advantage and technical lead will be challenging. The success of its open-source ecosystem depends on developer adoption and collaboration, which might not materialize as quickly or broadly as anticipated, potentially limiting its long-term competitive moat.

Market signals

00100· HKEX
  • 00100 The article reports a 15% stock surge for MiniMax-W following the H3 model announcement, indicating a positive market reaction.

AI-generated analysis of potential market relevance. Not financial advice.

Originally reported at

36kr.com

Discernion covers the story. Read the full piece at the source.

Tagschinaaillmstechbusinessstock-market

Author

Yinjian Finance

Intelligence analysis by

Gemini 2.5 Flash

Published

Jul 31, 2026

Source

36kr.com

Share

Topics

chinaaillmstechbusinessstock-market

Related

More from this desk

Jul 30·scmp.com

In 2016, Pokemon Go fever sweeps Hong Kong as players hunt citywide

Hong Kong became the second place in Asia to get Pokemon Go, sparking a citywide craze and trespassing warnings.

Why Andrew Ng Invested in LearnVector After 15 Years: Coursera's $100M Bet on AI-Driven Learning

Jul 30·36kr.com

Why Andrew Ng Invested in LearnVector After 15 Years: Coursera's $100M Bet on AI-Driven Learning

Andrew Ng, co-founder of Coursera, has invested in LearnVector, a new AI-driven learning platform. Coursera has invested $100M in LearnVector, acquiring a 33% stake. LearnVector aims to provide personalized learning experiences using AI, focusing on knowledge workers, ent…

Microsoft Earns $216 Billion Thanks to Anthropic

Jul 30·36kr.com

Microsoft Earns $216 Billion Thanks to Anthropic

Microsoft's AI business continues to grow, with Azure's quarterly revenue increasing by 43% and Microsoft 365 Copilot's paid seats exceeding 30 million. However, the company's capital expenditure has also increased, with a 70% year-over-year growth to $41 billion, putting…

AI Can't Jump: Why Large Language Models Can't Make Scientific Discoveries

Jul 30·36kr.com

AI Can't Jump: Why Large Language Models Can't Make Scientific Discoveries

A recent paper by Google DeepMind researchers suggests that large language models (LLMs) are unable to make scientific discoveries, despite their ability to process and analyze vast amounts of data. The researchers argue that LLMs lack the ability to make creative leaps, …