Tsinghua Professor Proposes "Moyuan" (模元) as Chinese Name for AI Token

Proposal

Yang Bin (杨斌), director of Tsinghua University’s Institute for Sustainable Social Value (清华大学可持续社会价值研究院), has proposed that the core AI measurement unit known as "token" be given a bespoke Chinese name: "模元" (Moyuan). The suggestion was published via Tencent Research Institute (腾讯研究院) and aims to distinguish AI tokens from other uses of "token" in blockchain and cybersecurity. It has been reported that Nvidia (英伟达) CEO Jensen Huang (黄仁勋) mentioned "token" more than 70 times at GTC 2026, underscoring the term's growing currency in global AI discourse.

Why "Moyuan"?

Yang argues that "模" (model/multi‑modal) anchors the term to the defining technology of the era — large models and multimodal systems — while "元" (element/unit) follows Chinese measurement naming logic such as 字节 (byte). He reportedly rejects pure transliterations like "托肯" or "屯" as lacking semantic clarity and being hard for non‑experts to accept. The proposed phrase is pitched as public‑friendly ("模元" feels less foreign than the English), industry‑useful (enabling terms like "模元消耗量", "模元效率"), and future‑compatible across agents, multimodal fusion and physical‑world AI.

Context and reaction

IT之家 ran a poll with 1,120 respondents offering several Chinese alternatives — including 模元, 词元, 语元, transliterations and other neologisms — reflecting how contested this small word has become. Naming debates matter beyond semantics: at a time when Nvidia GPUs and other AI hardware are political flashpoints in US‑China tech tensions and export controls, choices about terminology feed into localization, public understanding and industrial standardization. Will "moyuan" stick and become the shorthand of China’s AI era? That remains to be seen.