Elon Musk put a date on the debate on X, saying Chinese large language models could reach Anthropic’s Fable level in the first quarter of 2027. The response quickly drew a rebuttal from Zhipu AI founder Tang Jie, who answered with a short line: it will not take that long.
The exchange began with a question about when Chinese models might catch up with Anthropic’s Fable. Musk first gave the Q1 2027 estimate, then added a distinction. If the comparison is based on benchmarks, that timeline may fit. If the standard is real-world usefulness, even reaching that level by then would already be highly impressive. The discussion moved fast, and so did the metric being debated.
Musk shifts the focus from benchmark scores to revenue
Musk went on to say that Anthropic is correctly focused on maximizing “practical intelligence”. In his view, that quality may not show up in benchmark results, but it will show up in revenue. The point is simple: topping leaderboards and building something people will pay for are not the same thing.
That distinction gave the exchange a sharper edge. The question was no longer only about how quickly Chinese models can close a gap on paper. It also became a debate over product usefulness and commercial traction, two areas that do not always move in lockstep with benchmark performance.
Claude Fable 5 and GLM-5.2 frame the latest comparison
The timing of the comments matters because both sides have recently launched new models. Anthropic unveiled Claude Fable 5 on June 9. According to the source material, it belongs to the company’s most advanced Mythos-tier lineup and leads across many AI benchmarks. On SWE-bench Pro, a test tied to real engineering ability, it reportedly cleared 80%.
On the Chinese side, Zhipu released its flagship GLM-5.2 on June 17. The model highlights 1 million-token lossless long context, stronger coding capabilities, and support for domestic computing platforms on Day0. Tang Jie’s reply that it would take less time than Musk suggested was widely read as a direct expression of confidence in that progress.
The exchange did not settle the question, but it did clarify where the disagreement sits. Musk offered a benchmark-based timeline while stressing that practical utility will be reflected in revenue. Tang answered with a shorter and more aggressive timetable. The gap under debate is not just technical anymore; it is also about what counts as catching up.