ByteDance and DeepSeek Are Pursuing Divergent AI Strategies

Go high or go wide? DeepSeek and ByteDance, the two frontrunners in China’s AI sector, are taking dramatically different approaches. On Monday, DeepSeek launched DeepSeek V3.2, an open-weight model that invites experimentation. The startup claims it matches the performance of the latest offerings from OpenAI and Google, even surpassing them in specific mathematics benchmarks.

On that very day, ByteDance, which we previously discussed regarding its stronghold in AI applications, unveiled new features for users of its chatbot, Doubao. The company is now collaborating with a Chinese smartphone maker to integrate Doubao within the operating system, giving it the ability to access various apps and perform interactive tasks. In essence, it’s gearing up to challenge Apple’s Siri.

Both ByteDance and DeepSeek boast AI applications with over 140 million monthly users. However, their recent announcements highlight two diverging paths in China’s AI landscape. While certain firms continue to compete with their Western counterparts in developing increasingly sophisticated models, others have subtly retreated from that competition, focusing instead on embedding their AI technologies into daily life.

DeepSeek Resurfaces

DeepSeek’s newest open-weight model may have left some of its devoted followers underwhelmed, as many eagerly await R2, a highly awaited upgrade to the original model that made waves in Silicon Valley back in January. Instead, DeepSeek showcased V3.2 and V3.2-Speciale, optimized versions of its prior model, V3.2-Exp, launched in September.

Nonetheless, V3.2 generated excitement in the AI community because DeepSeek asserts it can tackle complex math questions posed at the International Mathematical Olympiad, while its performance in various coding and reasoning tasks reportedly meets or exceeds those of GPT 5 and Gemini 3. “It suddenly dawned on me why they call the company DeepSeek with the whale as a motif. Because just like a whale, it rarely surfaces, but every time it does, it creates a significant wave,” reflects Jen Zhu Scott, an AI investor and cofounder and CEO of Power Dynamics, a modular data-center solutions company.

However, there’s a sense that this hurried race to roll out new AI models is becoming somewhat wearisome, especially given the number of releases in the past month, each claiming to elevate humanity another notch. In just under 20 days, we witnessed OpenAI’s GPT 5.1, Google’s Gemini 3 Pro, and Anthropic’s Claude Opus 4.5; alongside Chinese open-source models like Moonshot’s Kimi K2 and DeepSeek’s V3.2, the landscape turns chaotic. My attention span can be summed up by this perfect meme.

“At the end of the day, we can’t keep up with all these fine differences between various models, different releases,” Zhu adds. “It actually doesn’t make a significant difference, apart from some sort of stock market speculation on which one will come out on top.”