On June 18, in response to a discussion on Twitter/X about when Chinese models would catch up to the capability level of Anthropic Fable (Mythos), Tang Jie (@jietang), founder of Zhipu z.ai and professor at Tsinghua University, replied with a brief „won’t take that long“ to Elon Musk’s earlier prediction of „probably Q1 (first quarter of 2027)“, hinting that the timeline would be sooner than market expectations. The discussion originated from an analysis post by the AI observation account Teortaxes: it argued that GLM-5.2 had already reached the level of Claude Opus 4.7-4.8, with a gap of about 7 months from the top models, and thus estimated that domestic „Fable-level“ models would emerge between November and December 2026. Musk then added in a follow-up comment that catching up on benchmark scores would „also be very impressive if it happens in Q1“, but stressed that Anthropic focuses on maximizing „real usefulness“ rather than benchmark scores, which „isn’t reflected in the scores, but is clearly reflected in revenue.“ Tang Jie replied, „All we need is focus, especially on understanding the nature of intelligence.“