CEO of Z.ai Responds to Musk: China Will Have Mythos-Level AI Sooner Than Expected

What Happened?

Elon Musk, CEO of xAI and a central figure in the artificial intelligence debate, recently stated that China will likely have a large language model (LLM) with 'Mythos'-level capabilities — the highest tier in Fable's classification, which includes levels like Fable 1-5 and Mythos — by the first quarter of 2027. However, Jie Tang, CEO of Chinese startup Z.ai (considered a rival to Anthropic), directly responded to Musk on social media: 'It won't take that long', suggesting his company will reach that milestone sooner than expected. Tang did not provide a specific date, but his statement has generated expectations of a potential imminent breakthrough. This exchange, initially reported by Tom's Hardware, reflects the growing competition in the race for artificial general intelligence (AGI).

Why Is This Important?

Mythos-level AI represents a threshold of general or near-general artificial intelligence, where models far surpass human capabilities in cognitive tasks. So far, no public model has reached that level. Tang's claim suggests that Z.ai may be close to a significant breakthrough, which would have geopolitical and technological implications. China has already demonstrated competitive capabilities with models like DeepSeek-V3, which competes with GPT-4 on several benchmarks, but achieving Mythos would represent a qualitative leap. The response also reflects the growing rivalry between Chinese and American startups, such as Anthropic (creator of Claude) and OpenAI. According to PitchBook data, AI investment in China reached $78 billion in 2024, surpassing the United States for the first time in AI startup funding.

Context and Background

Musk has been a vocal critic of AI risks and has promoted his own company, xAI, with the Grok chatbot. His prediction about China is set against a backdrop of technological acceleration: countries like China are heavily investing in AI, and companies like Z.ai are seeking to position themselves as leaders. The 'Fable' classification is an internal metric used by some organizations to measure progress toward AGI; Mythos would be the highest level, equivalent to a superintelligent AI. It is worth recalling that in 2023, Musk signed an open letter calling for a pause in training models more powerful than GPT-4, which contrasts with his current prediction that China will achieve Mythos in 2027. Meanwhile, Z.ai was founded in 2023 by Jie Tang, a former Baidu researcher, and has raised over $200 million in funding, according to Crunchbase. The company focuses on developing language models for enterprise applications, with an emphasis on computational efficiency.

What Will Be the Consequences?

If Z.ai achieves a Mythos model before 2027, it could redefine the global competitive balance. Companies like Anthropic and OpenAI could be pressured to accelerate their own developments. Additionally, Tang's response could be a strategic move to attract investment and talent, or a genuine sign of technical progress. However, independent confirmation is lacking; until Z.ai publishes results or demos, claims should be taken with caution. Historically, similar promises have proven exaggerated: in 2020, OpenAI claimed GPT-3 was close to AGI, but today it is recognized as far from that level. If Z.ai delivers, it could trigger a new wave of investment in Chinese AI startups, as well as stricter trade restrictions from the United States, similar to sanctions on Huawei. On the other hand, if it fails to achieve the milestone, it could damage its credibility and trust in the Chinese AI ecosystem.

What Should Readers Know?

• Musk's prediction is not official or binding; it is a personal opinion based on his knowledge of the sector.
• Z.ai is a relatively young and little-known startup; its claim has not been verified by independent sources.
• The Mythos level does not have a universally accepted definition; each company may use its own scale, making comparisons difficult.
• The actual timeline will depend on advances in hardware (such as Nvidia chips subject to export restrictions), data, and algorithms, as well as geopolitical factors.
• The scientific community remains divided on what exactly constitutes AGI; some experts, like Yann LeCun, consider current LLMs far from that level.

Analysis and Perspectives

Tang's statement can be interpreted as an attempt to capitalize on media buzz. However, it also reflects the confidence of Chinese researchers in their ability to innovate. Historically, China has achieved rapid advances in areas like facial recognition and natural language processing, but achieving Mythos would require overcoming enormous technical challenges, such as computational efficiency and training data quality. Additionally, U.S. export restrictions on advanced chips, such as Nvidia's H100, could slow Chinese progress. Nevertheless, companies like Z.ai may be developing alternative hardware or more efficient training techniques. The international community will closely watch Z.ai's and other Chinese companies' next moves, as well as potential announcements at conferences like the World AI Conference in Shanghai. In parallel, Musk has proposed that xAI also seeks to achieve AGI by 2029, adding another layer of competition.

'It won't take that long' — Jie Tang, CEO of Z.ai, responding to Elon Musk.

Conclusion

Tang's response to Musk adds tension to the race for advanced AI. While Musk bets on a relatively conservative timeline (2027), Tang suggests the milestone could be reached in months or perhaps a year. The veracity of these claims will only be known over time. For now, the exchange underscores the intense competition and high expectations surrounding Mythos-level AI. Investors and regulators should prepare for scenarios where China leads in AGI, which could reshape the global technological landscape. As always, caution is key until concrete evidence is presented.