Elon Musk replied 'Probably Q1' on X to speculation about when Chinese AI models might reach capabilities comparable to Anthropic's Fable series, which launched around June 9, 2026. Z.ai responded with acceptance emojis, framing the exchange as a public challenge for their GLM models. Jie Tang's reported comment that the milestone will arrive sooner lacks independent confirmation in current posts or sources.
Well, there you have it folks. Elon expects a Chinese Fable class model by Q1 2027, and @Zai_org accepts the challenge. This is in line with Anthropic's estimate too. Q3-Q4 will be very busy.
@elonmusk @teortaxesTex ✍️✍️✍️
Contemporaneous discussion describes Fable models as consumer versions of Mythos-class systems with added guardrails, yet no precise benchmarks or capability thresholds have been tied to the term in this thread.
The Beijing-based lab behind the open-source GLM family has not announced a specific model version, performance target, or release window connected to the Musk timeline.
Many users expressed excitement about signals of Chinese fable-class AI models arriving by late 2026 or Q1 2027 because of rapid global competition toward AGI, while a few worried about economic harm to the US or criticized Elon Musk.
No Digg Deeper questions have been answered for this story yet.
Professor is serious
@elonmusk @teortaxesTex won’t take that long
OH SNAP The founder of Zhipu thinks the first Chinese model in the Fable class could be reached before the end of 2026. He probably means his model. He will certainly try to make that the case.
@elonmusk @teortaxesTex won’t take that long
@elonmusk @teortaxesTex won’t take that long
@flyingpetal472 @elonmusk @teortaxesTex add oil
@jietang @elonmusk @teortaxesTex 我将尊您为杰神
@jietang @elonmusk @teortaxesTex
@teortaxesTex @Zai_org Isn't that a bit too late for them? They'd better hurry, as I expect US with Mythos will try to sabotage any large training run happening in China very soon. So it might be either soon or never.
@teortaxesTex @Zai_org when they get rubin - there will be unmatched speed for agi, intelligence and till then benchmarks look diluted. esp - usa 4 ai giants(u know)
(chineese are banned from rubin btw)
@jietang @elonmusk @teortaxesTex Insane Aura
@teortaxesTex @Zai_org I actually don’t think so. Fable level model possibly requires scale, the big model smell, Chinese lab can reliably handle training model that size in the short term future
@TotallyNotParth @Zai_org it'll be nerfed on offensive capabilities
@jietang @elonmusk @teortaxesTex 😎
@jpFromTlon @Zai_org They will only start building big clusters this fall they have never trained even a 1T model this is already aggressive So, yes. The US will have a window of opportunity only possibly restrained by fear of kinetic retaliation
@teortaxesTex @Zai_org Elon is reading all your disrespect for Grok😁
@teortaxesTex @Zai_org very possible, but i wonder how it'll go really. would they stop releasing weights for these models? seems kinda dangerous no?
@jietang @elonmusk @teortaxesTex 冲冲冲🫡！
@teortaxesTex How long till they get blacklisted or even worse, OSS models outlawed in US
@RedRibbon131420 they are on the entity list Americans will find a way to disappoint
@teortaxesTex @Zai_org If this gap persists, institutions would technically have six months to harden their infrastructure against potential cybersecurity threats..which isn’t much time.
@teortaxesTex @Zai_org who is that guy？
