Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Go high or wide? DeepSeek and ByteDance, the leaders Artificial intelligence industry in ChinaThey adopt vastly different strategies.
On Monday, DeepSeek released DeepSeek V3.2, another piece of software Open weight model That anyone can tamper with. The startup says it works on par with the latest models from OpenAI and Google, and even beats them on some basic math benchmarks.
On the same day, ByteDance, from Dominance in artificial intelligence applications We’ve previously covered and presented More than that Ways for people to use Doubao chatbot. ByteDance is now working with a Chinese smartphone maker to embed Doubao into the operating system, giving it access to various apps and allowing it to perform proxy tasks with them. In other words, it’s coming for Apple’s Siri.
ByteDance and DeepSeek both have AI applications with more than 140 million monthly users. But their recent announcements represent two diverging trends in China’s AI industry. While some companies are still competing with their Western counterparts to build more capable models, others have quietly withdrawn from that game and are focusing on how to integrate their AI tools into people’s daily lives.
The re-emergence of DeepSeek
DeepSeek’s latest open-weight prototype may have disappointed some of its most loyal followers, who are still waiting for R2, the long-awaited update to the prototype that rocked Silicon Valley in January. Instead, DeepSeek has released V3.2 and V3.2-Speciale, which are better-optimized versions of its previous model, V3.2-Exp, which was released in September.
However, version 3.2 has created a buzz in the AI industry because DeepSeek claims it can solve the kind of advanced mathematics questions asked at the International Maths Olympiad, and its performance on other coding and reasoning tasks is supposed to be equal to or higher than GPT 5 and Gemini 3. “It suddenly occurred to me why they call DeepSeek a whale as an idea. Because just like a whale, it rarely surfaces, but every time it surfaces, it always makes a big noise,” Jin says. Chu Scott, an AI investor and co-founder and CEO of Power Dynamics, a modular data center solutions company.
However, I can’t help but feel that the AI model arms race is getting a bit tiresome, especially since so many new models have been released in the past month, each claiming to take humanity a step up. In less than 20 days, we had OpenAI’s GPT 5.1, Google’s Gemini 3 Pro, and Anthropic’s Claude Opus 4.5; When using open source Chinese models like Moonshot’s Kimi K2 and DeepSeek’s V3.2, it becomes a complete mess. My attention span can be summed up by This is the perfect meme.
“Ultimately, we can’t keep up with all these subtle differences between different models and different versions,” says Chu. “It actually doesn’t make much difference, apart from some sort of stock market speculation about who’s going to win.”