MiniMax eyes Shanghai listing after Hong Kong success

🔑 Enhanced Key Takeaways

•MiniMax officially listed on the Hong Kong Stock Exchange in January 2026, raising approximately $619 million at an initial valuation near $6.5 billion.
•The company has secured approximately $1.15 billion in total funding from a diverse group of investors, including Alibaba, Tencent, Hillhouse Investment, and the gaming studio MiHoYo.
•MiniMax's annual recurring revenue (ARR) surpassed $300 million as of May 2026, with its 2025 revenue reported at $79.038 million, primarily driven by its AI-native products.
•MiniMax recently unveiled its M3 flagship AI model, designed for coding agents and automated workflows, which can process up to 1 million tokens and reportedly outperformed OpenAI's GPT-5.5 and Google's Gemini 3.1 Pro on the SWE-Bench Pro coding benchmark.

📊 Competitor Analysis▸ Show

Feature/Metric	MiniMax	Moonshot AI (Kimi)	Zhipu AI (GLM)	DeepSeek	ByteDance (Doubao)	OpenAI (GPT-5.5)	Google (Gemini 3.1 Pro)	Anthropic (Claude Opus 4.7)
Model Type	Multimodal LLMs (M3, M2.x, Hailuo, Speech, Music)	LLMs (Kimi K2.x)	LLMs (GLM series)	LLMs (R1)	LLMs (Doubao)	LLMs (GPT-5.5)	LLMs (Gemini 3.1 Pro, 2.5 Pro)	LLMs (Claude Sonnet 4.5, Opus 4.7)
Key Products	Talkie (AI character app), Hailuo AI (video gen), MiniMax Agent, M3 (coding/agentic)	Kimi Chat	-	-	-	ChatGPT, APIs	Gemini, APIs	Claude, APIs
Pricing (per 1M tokens)	M2.5: $0.15 (input), $1.20 (output)	-	-	-	$0.02 (API)	-	-	-
SWE-Bench Pro	M3: Beat GPT-5.5 and Gemini 3.1 Pro	-	-	-	-	M3 beat it	M3 beat it	-
Artificial Analysis Intelligence Index	M2: Ahead of Gemini 2.5 Pro	Kimi K2.6: 54	-	-	-	GPT-5.5: 60	Gemini 3.1 Pro: 57, Gemini 2.5 Pro: Below M2	Claude Opus 4.7: 57, Claude Sonnet 4.5: Outperformed by Kimi K2 Thinking
MLE Bench Lite	M2.7: Tied Gemini 3.1 Pro	-	-	-	-	-	Tied by M2.7	-

🛠️ Technical Deep Dive

MiniMax develops multimodal AI models capable of understanding, generating, and integrating text, audio, image, video, and music.
Flagship models include MiniMax M3, Hailuo 2.3 (video generation), Speech 2.8, and Music 2.6.
The recently unveiled MiniMax M3 is a frontier coding and agentic model built on a novel attention architecture (MSA) with a 1 million token context window.
M3 utilizes a MiniMax Sparse Attention mechanism, which reportedly increases inference speed by 9.7x during the prefilling phase (for 1 million tokens) and up to 15.6x faster decoding with a 1 million KV length, built on the GQA (Grouped Query Attention) architecture.
The MiniMax M2 series (M2, M2.5, M2.7) employs a sparse Mixture-of-Experts (MoE) decoder-only Transformer architecture.
MiniMax M2.7 is a 230-billion-parameter text-to-text AI model that incorporates a recursive self-optimization process, allowing it to generate its own evaluation data and synthetic training examples to improve performance by approximately 30% on internal benchmarks.
M2.7 also features an "interleaved thinking" protocol for agentic workflows, where the model integrates its full thinking history into the conversation context.
MiniMax-Text-01 is a large language model with 456 billion parameters and a 4 million token context window.
The Hailuo AI Video Model is powered by proprietary model series including MiniMax Hailuo 2.3 Fast, MiniMax Hailuo 2.3, and MiniMax Hailuo 02.
The Speech-02 model converts text to speech in 40 languages, can process up to 200,000 characters, and is ranked #1 on the Artificial Analyze Text-to-Speech leaderboard.

🔮 Future ImplicationsAI analysis grounded in cited sources

MiniMax's dual listing strategy will significantly enhance its capital-raising capabilities and market influence.

Listing on both Hong Kong and Shanghai's STAR Market provides access to broader investor pools (international and domestic) and aligns with China's strategic investment in critical technologies, potentially boosting its valuation and funding for costly AI development.

The company's focus on coding agents and automated workflows with models like M3 will accelerate its commercialization and enterprise adoption.

Coding assistants and autonomous agents are widely considered among the most commercially promising applications of generative AI, and M3's reported benchmark performance in this area positions MiniMax strongly in a competitive enterprise market.

MiniMax's recursive self-optimization framework (as seen in M2.7) could set a new standard for AI model development, leading to faster and more cost-effective improvements.

This innovative approach allows models to identify and address their own performance gaps without relying solely on expensive external retraining cycles, potentially reducing development costs and accelerating the pace of AI innovation.

⏳ Timeline

2021-12

MiniMax founded in Shanghai by former SenseTime researchers.

2022-10

Launched Glow (later Talkie), an AI character app.

2024-03

Alibaba Group led a $600 million financing round, valuing MiniMax at $2.5 billion.

2025-07

Closed a $300 million Series B extension at a $4 billion valuation, with Shanghai state-owned capital investment.

2026-01

Listed on the Hong Kong Stock Exchange, raising approximately $619 million.

2026-05

Initiated IPO tutoring process for a secondary listing on Shanghai's STAR Market.

MiniMax eyes Shanghai listing after Hong Kong success

⚡ 30-Second TL;DR

Key Points

🧠 Deep Insight

🔑 Enhanced Key Takeaways

🛠️ Technical Deep Dive

🔮 Future ImplicationsAI analysis grounded in cited sources

⏳ Timeline

📎 Sources (18)

👉Related Updates

IBM: AI spending is delaying, not killing, software deals