🏠IT之家•Stalecollected in 11m
MiniMax Officially Starts A-Share IPO Process
💡Major Chinese AI model developer MiniMax is expanding its capital footprint with an A-share IPO.
⚡ 30-Second TL;DR
What Changed
Signed IPO tutoring agreement with CITIC Securities
Why It Matters
MiniMax's dual-market listing strategy highlights the growing capital interest in Chinese AI infrastructure and model development companies.
What To Do Next
Monitor MiniMax's public financial disclosures for insights into their R&D spending and infrastructure scaling strategies.
Who should care:Founders & Product Leaders
Key Points
- •Signed IPO tutoring agreement with CITIC Securities
- •Company is a major player in the Chinese AI large model market
- •Previously listed on the Hong Kong Stock Exchange in January
🧠 Deep Insight
Web-grounded analysis with 22 cited sources.
🔑 Enhanced Key Takeaways
- •MiniMax Group officially listed on the Main Board of the Hong Kong Stock Exchange on January 9, 2026, under the stock code 0100.
- •The Hong Kong IPO successfully raised approximately HK$4.82 billion (US$620 million) by issuing 29.2 million shares at HK$165 per share.
- •The IPO was met with significant market enthusiasm, with the retail offering being oversubscribed an astonishing 1,837 times, setting a new record for recent new share issues in the Hong Kong market.
- •During its Hong Kong IPO, MiniMax Group was valued between approximately $5.93 billion and $6.47 billion.
- •The company plans to allocate approximately 90% of the IPO proceeds over the next five years towards research and development, specifically for large language model development and AI-native product development.
📊 Competitor Analysis▸ Show
| Feature/Metric | MiniMax M2.5 | DeepSeek-R1 | Anthropic Claude Opus 4.6 |
|---|---|---|---|
| Input Token Price (per 1M) | $0.15 - $0.30 | ~$0.02 | $5.00 |
| Output Token Price (per 1M) | $1.20 - $2.40 | N/A (generally lower) | N/A (generally higher) |
| SWE-Bench Verified Score | 80.2% | On par with GPT-4, Llama 3.1, Claude | 80.8% |
| Context Window | 204,800 tokens | N/A | N/A |
| Key Capabilities | Real-world productivity, coding, office tasks, agentic workflows | General-purpose LLM, cost-effective | Advanced reasoning, coding, multi-step tasks |
🛠️ Technical Deep Dive
- MiniMax develops multimodal AI models capable of processing text, audio, image, video, and music.
- Their model lineup includes the MiniMax M series (M2, M2.5, M2.7), Hailuo (video generation), Speech (text-to-speech), and Music.
- MiniMax M2.7 is a 230-billion-parameter Mixture-of-Experts (MoE) model with 10 billion active parameters and 62 layers.
- The M2.7 model incorporates a recursive self-optimization framework, allowing it to generate its own evaluation data, identify capability gaps, and produce synthetic training examples to improve performance, reportedly by 30% on internal benchmarks.
- An earlier model, MiniMax-01, was a hybrid MoE model featuring 456 billion total parameters and approximately 45.9 billion active parameters per forward pass, utilizing a combination of "Lightning Attention" for long-range context and standard softmax attention for local precision.
- MiniMax is actively developing the M3 series, which is expected to feature a new sparse attention mechanism designed to achieve up to 15.6 times faster decoding speed at 1 million tokens and a 9.7 times speedup in prefilling latency.
- The M2.5 model has achieved scores of 80.2% on SWE-Bench Verified, 51.3% on Multi-SWE-Bench, and 76.3% on BrowseComp.
- The M2.7 model demonstrates strong performance on various benchmarks, including 56.22% on SWE-Pro, 55.6% on VIBE-Pro, 57.0% on Terminal Bench 2, and an ELO score of 1495 on GDPval-AA.
🔮 Future ImplicationsAI analysis grounded in cited sources
MiniMax Group will significantly accelerate its AI model and product development.
The allocation of approximately 90% of its Hong Kong IPO proceeds to R&D over the next five years indicates a strong commitment to technological advancement and expansion.
The company's A-share IPO process will likely bolster its position in the domestic Chinese AI market.
Pursuing an A-share listing in mainland China, following its Hong Kong debut, suggests a strategic move to tap into domestic capital and strengthen its presence within China's rapidly growing AI ecosystem.
MiniMax's focus on self-optimizing models and efficient long-context processing will lead to more advanced AI agent capabilities.
The technical advancements in M2.7's self-optimization and the teased M3's sparse attention for long-context inference are geared towards creating more autonomous and cost-effective AI agents.
⏳ Timeline
2021-12
MiniMax Group founded in Shanghai.
2022-10
Launched its first product, Glow, an AI companion app.
2023-06
Launched Talkie, the international counterpart to Glow, for overseas markets.
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, reaching a $4 billion valuation.
2025-12-31
MiniMax officially launched its Initial Public Offering (IPO) in Hong Kong.
2026-01-09
MiniMax listed on the Hong Kong Stock Exchange under stock code 0100.
2026-03-18
MiniMax M2.7 model released.
2026-05-27
Released a technical report on M2 series models and teased the upcoming M3 series.
📎 Sources (22)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
📰
Weekly AI Recap
Read this week's curated digest of top AI events →
👉Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: IT之家 ↗