Go high or wide? DeepSeek and ByteDance, two leaders in China’s artificial intelligence industry, are adopting completely different strategies.
On Monday, DeepSeek released the DeepSeek V3.2, another open-weight model that anyone can tinker with. The startup claims it performs on par with the latest OpenAI and Google models, and even beats them in some key math tests.
On the same day, ByteDance, whose dominance in AI applications we discussed earlier, was presented even more ways to apply the Doubao chatbot. ByteDance is currently working with the Chinese smartphone manufacturer to embed Doubao into the operating system, giving it access to various applications and enabling it to perform agent tasks using them. In other words, Apple’s Siri is coming.
Both ByteDance and DeepSeek have AI applications with over 140 million monthly users. However, their latest announcements represent two divergent trends in China’s AI industry. While some companies continue to compete with their Western counterparts to build increasingly capable models, others have quietly withdrawn from the game and are focusing on how they can integrate their artificial intelligence tools into people’s everyday lives.
DeepSeek returns to the surface
DeepSeek’s latest open-weight model may have disappointed some of its most committed supporters, who are still waiting for R2, the long-awaited update to the original model that rocked Silicon Valley in January. Instead, DeepSeek released V3.2 and V3.2-Speciale, which are better optimized versions of the previous V3.2-Exp model, released in September.
Still, version 3.2 has caused a stir in the AI industry because DeepSeek claims it can solve the type of advanced math questions asked in the International Mathematical Olympiad, and its performance on other coding and reasoning tasks is allegedly comparable to or better than GPT 5 and Gemini 3. “Suddenly it dawned on me why they call DeepSeek a whale-themed company. Because, like a whale, it rarely comes to the surface, but every time it does, it always comes to the surface. makes a huge splash,” says Jen Zhu Scott, an artificial intelligence investor and co-founder and CEO of Power Dynamics, a company that offers modular data center solutions.
But I can’t help but feel that this arms race of AI models is getting a little tiring, especially since so many new ones were released last month, each one designed to take humanity a step further. In less than 20 days we had OpenAI’s GPT 5.1, Google’s Gemini 3 Pro, Anthropic’s Claude Opus 4.5; throw in Chinese open source models like Moonshot’s Kimi K2 and DeepSeek’s V3.2, and it’s a total mess. My focus can be summarized as follows: this perfect meme.
“Ultimately, we can’t keep up with all these fundamental differences between different models and releases,” Zhu says. “It doesn’t really make much of a difference, other than some sort of stock speculation about who’s going to win.”