Go high or go wide? DeepSeek and ByteDance, the two leaders of China’s AI industry, are adopting vastly different strategies.
On Monday, DeepSeek released DeepSeek V3.2, another open-weight model that anyone can tinker with. The startup says it performs on par with the latest models from OpenAI and Google, and it even beats them on some key mathematics benchmarks.
That same day, ByteDance, whose dominance in AI applications we covered previously, introduced even more ways for people to use its chatbot, Doubao. ByteDance is now working with a Chinese smartphone manufacturer to embed Doubao into the operating system, giving it access to different apps and allowing it to conduct agentic tasks with them. In other words, it’s coming for Apple’s Siri.
Both ByteDance and DeepSeek have AI apps with over 140 million monthly users. But their latest announcements represent two diverging trends in China’s AI industry. While some companies are still competing with their Western counterparts to build ever more capable models, others have quietly withdrawn from that game and are focusing on how they can integrate their AI tools into people’s everyday lives.
DeepSeek Resurfaces
DeepSeek’s latest open-weight model may have disappointed some of its most loyal followers, who are still waiting for R2, a much-anticipated update to the initial model that rocked Silicon Valley in January. Instead, DeepSeek released V3.2 and V3.2-Speciale, which are better-optimized versions of its previous model V3.2-Exp, released in September.
Still, V3.2 caused a stir in the AI industry because DeepSeek claims it can solve the type of advanced math questions asked at the International Mathematical Olympiad, and its performance on other coding and reasoning tasks is supposedly on par with or above 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 surfaces, it always makes a massive splash,” says Jen Zhu Scott, an AI investor and the cofounder and CEO of Power Dynamics, a modular data-center solutions firm.
However, I can’t help but feel like this arms race of AI models is getting a little tiring, particularly because so many new ones have been released in the last month, each claiming to take humanity one step higher. 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 becomes a total mess. My attention span can be summarized by this perfect meme.
“At the end of the day, we can’t keep up with all these hairline differences between different models, different releases,” Zhu says. “It actually doesn’t make a huge difference, apart from some kind of stock market speculation on who’s gonna win.”
DeepSeek, however, has differentiated itself in one important way: From the beginning, the company has focused on model efficiency, partly since it reportedly doesn’t have access to an abundant supply of computer chips. As a result, its models often use fewer computing resources to train and cost developers less to run than those from Western labs. “Efficiency is very important if you are a startup. An open source model is superior both in terms of costs and because of customizability,” says Steve Hsu, a physics professor at Michigan State University and an AI startup founder.
Operating System Ambitions
Meanwhile, ByteDance is on a completely different trajectory. The social media giant is still not making smartphones, but it is turning its AI chatbot Doubao into something as close to a smartphone operating system as possible.
In late November, ByteDance released Doubao Input Method. Input apps (also called keyboard apps) are third-party tools that allow people to type Chinese characters quickly on QWERTY keyboards and are every Chinese person’s portal into the digital world. Now, Doubao is offering a new keyboard app with what it claims is superior speech-to-text functionality, giving users another entry point into ByteDance’s AI app ecosystem.
But ByteDance’s most ambitious move recently came on Monday, when it released a Doubao AI agent that can be integrated into a smartphone’s operating system, giving it control over any app. In a preview video, ByteDance shows how Doubao can access Tesla’s app and open the trunk using voice inputs, search through different ecommerce platforms to find the lowest prices, and access photos in a user’s camera roll and enhance them with AI.
ByteDance is working with the Chinese smartphone manufacturer ZTE to preinstall the Doubao agent on one of its phone models, the Nubia M153, which sells for 3,499 RMB (about $500). ByteDance says it’s also talking to other smartphone makers about installing its agent, but it seems unlikely that many will take it up on the offer—the most popular Chinese smartphone brands, like Huawei or Xiaomi, are all developing their own proprietary AI agents.
Doubao’s agent has already been blocked by the most popular app in China, Tencent’s WeChat. With over 1.4 billion users and an expansive range of features, WeChat is as close to an operating system as you can get in China today. On Wednesday, users on Chinese social media reported that they were suspended from WeChat after they used Doubao’s agent to access it, a punishment no one would be willing to risk receiving in China. Doubao quickly announced that it has disabled its agent from working with WeChat, and said that the suspended users will get their accounts back soon. On this latest episode of Chinese big-tech feuds, it’s Tencent 1 vs. ByteDance 0.
Where Chinese AI Is Heading
DeepSeek and ByteDance’s Doubao represent two ends of the spectrum of Chinese AI companies, while others fall somewhere in-between. Small but mighty startups like Zhipu, Minimax, and Moonshot are following in DeepSeek’s footsteps, releasing capable open source models, while Baidu and Tencent are going the more application-focused route. Alibaba is still mostly on DeepSeek’s side, constantly releasing new versions of its open source Qwen models, but lately, it seems eager to cross over to the other side: Last month, it put out a consumer-facing AI super-app.
ByteDance doesn’t have to compete in the benchmark arms race because it already has a gigantic user base, and as a private company, Hsu says, it doesn’t need to worry about stock market volatility. “They just have to ship quietly, integrate a really powerful AI model into their existing app, and I think that’s what they are doing,” he explains. Alibaba and Google could go that route, too, but they seem more hung up on prestige and winning accolades than ByteDance.
What unites Chinese AI firms is the thing they’re not doing. Unlike American tech giants, they’re steering clear of the compute hoarding game, where companies keep building more data centers and brute-forcing their way to more powerful models. Partly, that’s because of American chip sanctions, which block Chinese companies from accessing cutting-edge chips from Nvidia. But also, “they are just not as well capitalized. They don’t have infinite amounts of money the way that American companies have,” Hsu says, so it makes sense for them to opt out from the beginning.
This is an edition of Zeyi Yang and Louise Matsakis’ Made in China newsletter. Read previous newsletters here.
The post ByteDance and DeepSeek Are Placing Very Different AI Bets appeared first on Wired.
