Tang Jie, the CEO of Chinese AI startup Zhipu, did not just correct Elon Musk on a timeline. He challenged the very definition of what “useful intelligence” means. The exchange, which played out publicly on social media, was brief.
But the implications for the US-China AI race are not. It started with Musk.
He replied to a post about China’s timeline to launch a rival to Anthropic’s new Fable 5 model. “Probably Q1,” Musk said. Tang Jie answered with three words: “won’t take that long.” That is the headline.
The real story is what followed. Musk shifted the argument.
He said benchmarks do not matter. Even if Chinese models match Fable 5 on paper, he argued, “true usefulness” is a different thing. He pointed to revenue.
Anthropic focuses on “maximizing useful intelligence,” Musk said, and that shows up in money. Tang Jie did not let that stand. He replied: “Focus is all we need, especially focusing on what intelligence truly is.” That sentence is the core of this story.
Tang Jie did not argue that Zhipu’s model is faster or cheaper. He argued that Zhipu’s definition of intelligence is different.
He did not accept Musk’s frame. He rejected the premise that revenue is the measure of usefulness. He did not say Zhipu is catching up.
He said Zhipu is running a different race. The technical data backs him up.
Zhipu’s latest model, GLM-5.2, performs on par with Anthropic’s Opus 4.7 to 4.8, according to state and international reports. Those models launched in April and May of this year. GLM-5.2 also consistently outperforms OpenAI’s GPT-5.5.
That is not a small thing. GPT-5.5 is the latest from the company that started the generative AI boom. Zhipu is beating it on benchmarks.
But Musk is not wrong about revenue. Anthropic’s models cost money to use.
They are built for businesses. They are designed to be useful in a commercial sense. Zhipu, by contrast, is a Chinese company operating in a different market.
Its definition of “useful” may not be the same as Anthropic’s. Tang Jie’s comment suggests Zhipu is thinking about intelligence as a concept, not just a product.
This matters because officials on both sides are watching. The AI race between the US and China is not just about who builds the biggest model. It is about who builds the model that actually works for their economy, their government, their people.
If Zhipu’s approach yields a model that is more adaptable, more efficient, or more aligned with Chinese needs, it could shift the balance. Musk’s “Q1” prediction may still be wrong. Tang Jie said it will not take that long.
If he is right, a Fable 5-class Chinese model could arrive by the end of this year. That would change the conversation entirely.
The debate would no longer be about catching up. It would be about who leads. For now, the exchange is a rare glimpse into how two very different AI leaders think.
Musk sees intelligence as a tool to be sold. Tang Jie sees it as something to be understood.
Both are probably right. But only one of them is betting on speed.
Sources
- Zhipu Founder Tang Jie's Remote Dialogue with Elon Musk – 36氪
- Elon Musk Says Chinese LLMs Will Reach Fable Level… – EastFrontier
