Zhipu's GLM-5.2 Ranks No. 2 Globally; Tang Disputes Musk Timeline

Zhipu AI released its GLM-5.2 model on June 16 and within days it had landed the No. 2 position on Code Arena's front-end coding benchmark, behind only Anthropic's Fable 5, with an Elo score of 1595. The detail that elevates this beyond a typical leaderboard shuffle: the entire GLM-5 model family was trained on 100,000 Huawei Ascend 910B processors, with no Nvidia hardware involved at any stage.

That result set off a pointed exchange on X. According to the South China Morning Post, on June 18 Elon Musk projected that China would produce a Fable-class rival "probably" by the first quarter of next year. Tang Jie, founder and chief scientist of Zhipu AI, replied that it "won't take that long."

The market reaction was swift. On June 22, Zhipu AI's shares surged 42 per cent intraday to HK$2,980, carrying the company's market capitalisation past HK$1 trillion (roughly US$128 billion). GLM-5.2 had also ranked as the top open-weight model on Artificial Analysis's Intelligence Index, adding to a run of strong benchmark showings in recent weeks.

The honest caveat is that Code Arena's front-end coding sub-benchmark is a narrow slice of frontier capability, and an Elo score of 1595 in one category does not settle the broader question of parity across reasoning, safety, or multimodal tasks. Both Musk's "probably Q1" and Tang's "won't take that long" are informal declarations on a social platform, not product roadmaps, so take the timeline specifics as signals rather than schedules. What the reporting does not give you is a definition of what "matching Fable 5" actually means -- whether that is benchmark parity, latency parity, or commercial deployment readiness.

What is harder to dismiss is the hardware independence angle. A 744-billion-parameter model reaching No. 2 on a global coding benchmark without a single Nvidia chip is a meaningful proof point for China's independent AI supply chain, and it remains relevant regardless of whether the year-end parity claim ultimately holds.