MiniMax

MiniMax is a Chinese artificial intelligence company founded in 2021 by former SenseTime executives, developer of the M-series language models and the Hailuo video, Speech, and Music multimodal lines, and the second Chinese AI lab to complete a Hong Kong IPO.
MiniMax

MiniMax

MiniMax is a Chinese artificial intelligence company founded in 2021 by Yan Junjie, Yang Bin, and Zhou Yucong, all former senior researchers and engineers at SenseTime. The company is headquartered in Shanghai and develops a multimodal AI portfolio spanning the MiniMax M-series language models, the Hailuo video generation service, the Speech-02 audio model, the Music-01 music model, and the consumer companion-app products Talkie and Xingye. MiniMax completed its Hong Kong IPO on January 9, 2026, raising $619 million and reaching a market capitalization of approximately $11.5 billion at debut, making it the second pure-play foundation-model developer to go public after Z.ai / Zhipu AI.

At a glance

  • Founded: 2021 in Shanghai by Yan Junjie, Yang Bin, and Zhou Yucong (all former SenseTime).
  • Status: Public. Listed on the Hong Kong Stock Exchange since January 9, 2026.
  • Funding: Approximately $1.15 billion in private rounds prior to IPO. January 2026 IPO raised $619 million at debut. Pre-IPO investors include Alibaba, Tencent, MiHoYo, HongShan (Sequoia China successor), Hillhouse Investment, IDG Capital, and Shanghai State-owned Capital Investment.
  • CEO: Yan Junjie (co-founder; former senior executive at SenseTime).
  • Other notable leadership: Yang Bin (co-founder), Zhou Yucong (co-founder).
  • Open weights: Yes, partial. MiniMax M-series models including M2 have been released open-weights through Hugging Face. Earlier ABAB and MiniMax-VL-01 were closed-weights commercial releases.
  • Flagship models: MiniMax M2.7 (March 2026, top-tier on Artificial Analysis Intelligence Index), MiniMax M2.5 (February 2026), MiniMax M2 (open-weights, late 2025), Hailuo video generation, Speech-02 audio, Music-01.

Origins

MiniMax was founded in 2021 by Yan Junjie and two of his colleagues from SenseTime, the Hong Kong-listed Chinese computer-vision company that was one of the dominant Chinese AI organizations of the 2015 to 2020 deep-learning era. Yan had been a senior leader at SenseTime prior to founding MiniMax and brought a research-and-engineering background combining computer vision and large-scale model deployment.

The company's founding strategy combined consumer AI applications with foundation-model research. The Talkie character-companion app and the Xingye AI-character app provided consumer distribution and a continuous data feedback loop, while the underlying ABAB and later MiniMax-VL and M-series models supplied the model capability. The combination of consumer-application revenue and frontier-model research distinguished MiniMax from peer Chinese AI Tigers (Z.ai, Moonshot AI, 01.AI) that prioritized developer-and-enterprise distribution.

The 2024 to 2025 funding trajectory included a $250 million Series A led by Tencent in June 2023, a $600 million Series B led by Alibaba in March 2024 at a $2.5 billion valuation, and a $300 million round led by Shanghai State-owned Capital Investment in July 2025. Cumulative pre-IPO funding reached approximately $1.15 billion. The investor mix combined Chinese internet incumbents (Alibaba, Tencent), the gaming-and-content company MiHoYo (developer of Genshin Impact, an early Talkie commercial integration), and Chinese sovereign-and-state-affiliated capital.

The Hailuo video-generation service launched in 2024 and quickly established MiniMax as a leading Chinese video-generation provider, alongside Kuaishou's Kling, ByteDance's Seedance, and Alibaba's Wan. Hailuo-02 in 2025 expanded the capability profile and the consumer-and-enterprise distribution.

The Hong Kong IPO on January 9, 2026, raised $619 million and resulted in a same-day share-price gain of approximately 109 percent. The post-IPO market capitalization of approximately $11.5 billion placed MiniMax among the most highly valued Chinese AI Insurgents and the second pure-play foundation-model-developer public listing globally.

The flagship MiniMax M2.7 model launched on March 18, 2026, two months after the IPO. M2.7 reported the top position on the Artificial Analysis Intelligence Index at the time of its release, ranking first across the leaderboard's 136 evaluated models with a composite score of 50. The release reinforced MiniMax's competitive position alongside the leading Chinese open-and-closed-weights labs.

Mission and strategy

MiniMax's stated mission is to build artificial general intelligence with a focus on multimodal capability and on consumer-application distribution. The company has framed its strategic position as "Universal Intelligence for Everyone," with the consumer applications and the underlying multimodal foundation models as the dual pillars of the strategy.

The strategy combines four threads. First, frontier-tier foundation-model research, currently anchored by the M2.7 release on the Artificial Analysis leaderboard. Second, multimodal model breadth across text, video, audio, and music, with the Hailuo, Speech, and Music model lines providing capability differentiation beyond text. Third, consumer-application distribution through Talkie, Xingye, and Hailuo AI, building the daily-active-user base that competitors with developer-only distribution cannot match. Fourth, the public-company status post-January-2026, providing capital access for compute infrastructure and continued model investment.

The competitive premise is that the combination of frontier model capability, multimodal breadth, consumer-application distribution, and public-market capital produces a vertically integrated position that pure-research-stage Chinese AI Insurgents cannot match. MiniMax's "lightning attention" architectural innovation, a hybrid linear-and-softmax attention mechanism that reduces inference complexity from quadratic to linear, supports a 4-million-token context window at API pricing substantially below US frontier-lab equivalents.

Models and products

  • MiniMax M2.7. Released March 18, 2026. Current flagship in the M-series. Reported top position on the Artificial Analysis Intelligence Index at $0.30 per million input tokens, with reported scores of 56.2 percent on SWE-Pro and 57.0 percent on Terminal Bench 2.
  • MiniMax M2.5. Released February 2026. Multilingual coding-focused release reportedly performing on par with Claude Opus 4.5 at approximately $0.12 per million input tokens.
  • MiniMax M2. Released late 2025 as the first open-weights M-series flagship, with weights distributed through Hugging Face.
  • MiniMax M1. Earlier 2025 release.
  • MiniMax-VL-01 and MiniMax-01. December 2024 releases. Multimodal and text foundation-model variants.
  • ABAB family. 2024 closed-weights model line, the predecessor to the M-series.
  • Hailuo video generation (Hailuo-02 and prior). Video-generation model line. The flagship Hailuo product is one of the leading Chinese video-generation services for consumer and prosumer use.
  • Speech-02. Audio-generation and speech-synthesis model.
  • Music-01. Music-generation model.
  • Talkie and Xingye. AI-character consumer applications. The principal consumer distribution channels for MiniMax beyond the developer API.

The commercial channels span the MiniMax developer API platform, the consumer applications, and the Hugging Face open-weights distribution for the M-series.

Benchmarks and standing

MiniMax M2.7 reported the top position on the Artificial Analysis Intelligence Index at its March 2026 release with a composite score of 50, ranked first among 136 evaluated models. The benchmark profile combines frontier-tier capability with cost-efficient pricing ($0.30 per million input tokens at release).

On agentic-coding benchmarks, M2.7 reports 56.2 percent on SWE-Pro and 57.0 percent on Terminal Bench 2. M2.5 reports performance on par with Claude Opus 4.5 at substantially lower cost.

These positions are point-in-time and rotate with the release cadence of competing labs. The MiniMax capability profile is notable for combining frontier-tier benchmark performance with the multimodal breadth (Hailuo video, Speech-02, Music-01) that text-only competitors do not match. The lightning-attention architectural innovation is widely cited in industry coverage as a structural cost-efficiency advantage.

Leadership

As of April 2026, MiniMax's senior leadership includes:

  • Yan Junjie, Chief Executive Officer and co-founder. Former senior executive at SenseTime, with research and engineering background combining computer vision and large-scale model deployment. Public face for MiniMax's foundation-model research and IPO communications.
  • Yang Bin, co-founder.
  • Zhou Yucong, co-founder.

The company has hired aggressively from Chinese university research programs and from SenseTime, Alibaba, Tencent, and ByteDance. Specific senior research and engineering leadership beyond the three co-founders has been less broadly profiled in international media than for some peer Chinese AI Insurgents.

Funding and backers

MiniMax's funding history through April 2026 is approximately $1.77 billion cumulative, including $1.15 billion in private rounds and $619 million in the January 2026 IPO. Pre-IPO investors include Alibaba (Series B lead, March 2024), Tencent (Series A lead, June 2023), MiHoYo (the gaming-and-content company that integrated MiniMax models in its products), HongShan, Hillhouse Investment, IDG Capital, and Shanghai State-owned Capital Investment.

The post-IPO market capitalization of approximately $11.5 billion at debut and the 109 percent first-day share-price gain placed MiniMax among the most successful Chinese AI public-market debuts. The IPO followed Z.ai's January 8, 2026 listing by one day, with both companies forming the first wave of Chinese AI Tiger public listings.

The capital structure reflects diverse investor participation across Chinese internet incumbents, content-and-gaming companies, sovereign-and-state-affiliated funds, and venture capital. The Shanghai State-owned Capital lead in July 2025 was a structurally significant signal of state support.

Industry position

MiniMax occupies a structurally distinctive position among Chinese AI labs. The combination of consumer-application distribution at scale (Talkie, Xingye, Hailuo AI), multimodal model breadth (text, video, audio, music), public-company capital access (post-January-2026 IPO), and frontier-tier capability (M2.7's leaderboard position) produces a vertically integrated profile that no other Chinese AI Insurgent matches. Industry coverage has frequently characterized MiniMax as the most diversified Chinese AI Insurgent in terms of modality coverage and consumer-application distribution.

Strategic risks include the post-IPO public-company governance and disclosure requirements, intensifying competition on the foundation-model tier from DeepSeek, Alibaba Qwen, and Moonshot AI, intensifying competition on the video-generation tier from Kuaishou Kling and ByteDance Seedance, and the operational complexity of executing across multiple modalities and product lines simultaneously. Strategic strengths include the multimodal breadth, the consumer-application distribution, the public-company capital access, the founding-team SenseTime credentials, and the lightning-attention architectural innovation.

Competitive landscape

MiniMax competes with several Chinese and international AI labs:

  • DeepSeek, Alibaba Qwen, Moonshot AI, Z.ai / Zhipu AI. Direct Chinese frontier-model competitors. M2.7 competes with V4, Qwen 3.6, Kimi K2.6, and GLM-5 on capability and on the open-weights distribution market.
  • Baidu, Tencent Hunyuan, ByteDance Seed. Chinese incumbent AI competitors. MiniMax's distinguishing features against this cohort are the consumer-application focus and the public-company status.
  • Kuaishou Kling, Alibaba Wan. Direct competitors on Chinese video generation.
  • OpenAI Sora, Runway, Luma. International video-generation competitors.
  • Character.AI. International AI-character-companion competitor; Talkie and Xingye are MiniMax's analogs.
  • OpenAI, Anthropic, Google DeepMind. International frontier-model competitors. Less direct overlap on Chinese consumer distribution.

Outlook

Several open questions affect MiniMax's trajectory in 2026 and 2027:

  • The capability profile and release cadence of MiniMax M3 and successor models. M2.7 set a high benchmark bar; sustaining frontier-tier performance is the central technical question.
  • Hailuo video, Speech-02, and Music-01 successor releases. Multimodal capability differentiation remains a strategic priority.
  • The post-IPO commercial trajectory of Talkie, Xingye, Hailuo AI, and the developer API.
  • International expansion of consumer applications, particularly Talkie, against Character.AI and other international competitors.
  • Continued senior-talent recruitment, particularly at the post-IPO public-company scale.
  • US export-control developments affecting MiniMax's compute infrastructure and the broader Chinese AI hardware-and-software ecosystem.

Sources

About the author
Nextomoro

AI Research Lab Intelligence

Keep track of what's happening from cutting edge AI Research institutions.

AI Research Lab Intelligence

Great! You’ve successfully signed up.

Welcome back! You've successfully signed in.

You've successfully subscribed to AI Research Lab Intelligence.

Success! Check your email for magic link to sign-in.

Success! Your billing info has been updated.

Your billing was not updated.