MiniMax Joins China’s AI Startups in Open-Source Race
MiniMax stands out among China’s crop of AI startups pushing open-source models. Chinese open-source models now account for 30% of global AI use, per SCMP. WIRED groups it with Zhipu and Moonshot as “small but mighty” players following DeepSeek’s lead, putting out capable models anyone can tweak. These efforts come as Chinese firms split between raw model power and everyday apps.
Strong Backing for MiniMax and Peers
TokenRing calls MiniMax one of the new “AI Tigers,” alongside Baichuan AI, Moonshot AI, and Zhipu AI. These startups have raised billions in venture capital and hit multi-billion-dollar valuations. Zhipu AI and MiniMax plan IPOs that could raise billions more, Reuters reports. That’s amid a wave of investment in China’s AI scene, now over 4,500 companies strong with a market topping 700 billion yuan in 2024. Government backing aims for world leadership by 2030, but profitability lags due to price wars and open-source pressure.
For context, giants like Alibaba plan 380 billion yuan in spending over three years on AI compute. Yet startups like MiniMax face the same squeeze to turn tech into cash.
Read more on China’s AI Tigers in TokenRing
Performance Ties to Top Models
Specific benchmarks for MiniMax lag in these reports, but its path mirrors DeepSeek’s. WIRED notes DeepSeek’s V3.2 handles International Mathematical Olympiad-level math and matches or tops GPT-5 and Gemini 3 Pro on coding and reasoning, Indian Express reports. Released the same week as updates from OpenAI, Google, and Anthropic’s Claude, it stirred buzz for efficiency on limited chips.
- DeepSeek V3.2: Claims parity with GPT-5/Gemini 3 Pro overall, better on some math.
- MiniMax/Zhipu/Moonshot: Release similar open-source models, per WIRED.
No direct MiniMax numbers versus ChatGPT, Grok, or Claude here, but the group challenges Western leaders through cost savings and customization. DeepSeek, for one, trains leaner, helping developers run models cheaper, as Cybernews notes.
WIRED covers DeepSeek and MiniMax strategies
China’s AI firms dodge the U.S. compute arms race partly due to chip sanctions, focusing instead on open models and apps. MiniMax fits this shift, backed heavily but racing to profit.
