Why Kimi and MiniMax Dominate China | Generated by AI
Question: Why have Kimi (Moonshot AI) and MiniMax succeeded and beaten rivals like 01.AI, xAI (China’s context), Baichuan, and Wang Xiaochuan’s companies in the Chinese AI race?
Answer:
This is a great question about the competitive dynamics inside China’s AI industry. Here’s a comprehensive breakdown of why Kimi (Moonshot AI) and MiniMax have pulled ahead of earlier “AI tiger” rivals like Baichuan (Wang Xiaochuan), 01.AI (Li Kaifu), and others.
1. Superior Model Performance at Competitive Cost
MiniMax M2 hit 69.4% on SWE-Bench Verified, competing with models that cost 10x more. For a typical enterprise workload (100M input tokens + 20M output tokens monthly), MiniMax M2 costs just $24 — not $240, not $2,400. This extreme cost-efficiency is a decisive advantage.
Moonshot AI’s Kimi K2 Thinking outperformed OpenAI’s GPT-5 and Anthropic’s Claude Sonnet 4.5 on several benchmarks. Nathan Lambert of the Allen Institute for AI praised Kimi K2 Thinking for narrowing the gap between open-source models and the world’s leading closed-source systems.
2. Architectural Innovation (MoE)
MiniMax M2 adopts the Mixture-of-Experts (MoE) architecture with 200 billion parameters, but activates only 10 billion per forward pass, allowing frontier performance at dramatically lower compute cost.
MiniMax’s M2 runs efficiently on just 4×H100 GPUs and achieves 1,500 tokens per second — meaning your infrastructure costs drop dramatically.
3. Dominant Market Share in Token Usage
MiniMax M2.5 emerged as the most popular AI model by token usage on OpenRouter, with 4.55 trillion tokens used, and Kimi K2.5 ranked second with 4.02 trillion tokens — both far ahead of Western competitors in developer adoption.
Chinese models captured 61% of total token volume on OpenRouter, the world’s largest LLM API aggregation platform, with MiniMax and Kimi leading the pack.
4. Kimi’s Smart Ecosystem Moves (OpenClaw Strategy)
When the OpenClaw framework went viral in February 2026, Kimi moved before anyone else — it became the first model provider to offer free API credits to OpenClaw users. Moonshot also launched Kimi Claw, a one-click cloud deployment available to subscribers at 199 RMB/month, removing setup friction for non-technical users. OpenClaw alone accounted for roughly a quarter of all Kimi K2.5 overseas consumption.
This kind of rapid ecosystem integration is something Baichuan and 01.AI failed to execute.
5. Revenue Explosion for Kimi
In fewer than 20 days since the launch of Kimi K2.5, Kimi’s cumulative revenue already exceeded its total revenue for the entire calendar year 2025. Overseas revenue now surpasses domestic revenue for the first time.
6. MiniMax’s Global Consumer Success
MiniMax pivoted its early Glow concept to Talkie, an AI companion app targeting overseas markets. By June 2024, Talkie ranked as the 5th most-downloaded free entertainment app in the U.S. and surpassed Character.AI to become the No. 1 AI companion app in downloads, with over 11 million monthly active users worldwide.
MiniMax reported a better-than-expected 159% jump in annual revenue in 2025, with shares surging 21%, reflecting solid demand for low-cost, open-source Chinese models rivaling Silicon Valley.
7. Why the Others (Baichuan / 01.AI) Fell Behind
Baichuan (Wang Xiaochuan):
Baichuan dramatically changed direction. Two of its co-founders resigned to start their own business, just months after another top executive left the multibillion-dollar company. Baichuan pivoted heavily toward medical AI, narrowing its focus and ceding the general-purpose LLM race.
01.AI (Li Kaifu / Lee Kai-fu):
In August 2024, 01.AI lost its vice-president Huang Wenhao to ByteDance.
01.AI started focusing on enterprise-grade solutions powered by DeepSeek rather than building its own frontier models, effectively giving up the race to be an independent model leader.
8. The DeepSeek Price War Crushed Weaker Players
By May 2024, DeepSeek triggered a price war after launching its API at just $0.14 per million tokens — 5–10x lower than market rates. By January 2025, the average price of LLMs in China had fallen 92% compared to May 2024. Smaller companies without superior efficiency (like Baichuan and 01.AI) could not survive this compression.
9. China’s Talent Pool Advantage for the Winners
One AI industry executive noted that China’s deep talent pool was a key reason companies like Moonshot and MiniMax could punch above their weight, despite having far fewer high-end chips and less funding than many of their US peers.
Summary Table
| Company | Outcome | Key Reason |
|---|---|---|
| Kimi (Moonshot AI) | 🏆 Winner | Benchmark-beating models, ecosystem moves, revenue explosion |
| MiniMax | 🏆 Winner | MoE efficiency, global consumer apps, 159% revenue growth |
| Baichuan / Wang Xiaochuan | ❌ Declining | Leadership exodus, pivoted to medical niche |
| 01.AI / Li Kaifu | ❌ Declining | Key talent left, retreated to enterprise using others’ models |
| DeepSeek | 🥈 Competitor | Research/infra focus, declining consumer traffic |
In short: Kimi and MiniMax won because they combined frontier model quality, cost efficiency via MoE architecture, smart ecosystem partnerships, and global consumer product execution — while rivals like Baichuan and 01.AI lost talent, narrowed focus too early, and couldn’t survive the brutal AI price wars.
References:
- China’s Kimi Sets Record as Fastest Decacorn – China AI Daily
- Beyond DeepSeek: Moonshot and MiniMax step up – Yahoo Finance / SCMP
- MiniMax and Kimi Top Token Usage Rankings – SCMP
- China’s Open-Source AI Revolution: MiniMax M2 and Kimi K2 – Medium
- Kimi’s 20-Day Revenue Spike – Voice of Context
- Baichuan’s Big Pivot – The Wire China
- How DeepSeek Made China Investible Again – Deep Tech Asia
- Zhipu AI and MiniMax: Not China’s OpenAI – Recode China AI
- MiniMax Shares Soar After 159% Revenue Jump – Bloomberg