Zhipu AI founder Tang Jie told Elon Musk China will produce a rival to Anthropic's Fable 5 before the end of 2026. Musk had predicted the milestone would arrive in Q1 2027; Tang replied it 'won't take that long'.
Key facts
- GLM-5.2: 744 billion parameters.
- No. 2 on Code Arena front-end coding rankings.
- Zhipu market cap: HK$1 trillion (US$128 billion).
- Musk predicted Q1 2027; Tang says end of 2026.
- Zhipu ranked world's third-best AI lab by Artificial Analysis.
The exchange unfolded on X last week after Musk responded to a post about when China could match Fable 5 — currently the world's most advanced model According to SCMP. Tang's confidence is backed by real numbers: Zhipu's 744 billion-parameter GLM-5.2 recently grabbed the No. 2 spot globally on Code Arena's front-end coding rankings, trailing only Fable 5.
The benchmark evidence
Code Arena's front-end coding ranking places GLM-5.2 ahead of models from OpenAI, Google, and Meta. The model's performance, combined with Zhipu's rise to the world's third-best AI lab per the Artificial Analysis Intelligence Index, suggests the gap is narrowing faster than most Western analysts expected. Zhipu's market capitalisation hit HK$1 trillion (US$128 billion) on Monday, surging 1,700% from its IPO price, per SCMP.
The circular-debt subplot
Meanwhile, Micron invested in Anthropic's Series H round in exchange for a multi-year memory supply deal The Decoder reports. Anthropic co-founder Tom Brown called memory "critical to training and running Claude." Critics argue such circular deals inflate a bubble — Micron's stock has surged more than tenfold in a single year. The contrast is sharp: Zhipu is building frontier models with Chinese capital and homegrown silicon, while Anthropic locks in memory supply through equity-linked procurement.
What the timeline dispute reveals
Musk's Q1 2027 estimate is conservative by industry standards. The unique take here is not whether China will catch up, but that the catch-up is being driven by a single company — Zhipu — rather than the broader Chinese ecosystem. Zhipu's GLM-5.2 is the only Chinese model in Code Arena's top five. If Zhipu falters, the timeline slips. If it sustains this pace, Musk's estimate may prove too cautious.
Key Takeaways
- Zhipu's 744B-parameter GLM-5.2 ranks No.
- 2 globally on Code Arena.
- Tang Jie tells Musk China will match Fable 5 by end of 2026, not Q1 2027.
What to watch
Watch for the next Code Arena ranking update — if GLM-5.2 closes the gap to Fable 5 by more than 5 percentage points, Tang's timeline becomes credible. Also watch Zhipu's Q3 earnings for inference compute spend disclosure.
Source: scmp.com
