PeopleofSuperintelligence

Researcher

Aohan Zeng

Also known as 曾奥涵

Aohan Zeng researches language models, agent systems, and multimodal reasoning. He coauthored FastMoE in 2021, GLM-130B in 2022, AgentTuning in 2023, and LongBench and GLM-4-Voice in 2024. GLM-4.5 expanded his work to vision-language models and agentic reasoning in 2025; he also discussed smaller mixture-of-experts models at an August 2025 AMA. In 2026, he coauthored GLM-5 and IF-CRITIC; he is a Z.ai researcher and a computer-science PhD student at Tsinghua University.

Now at
Zhipu AI
Works
40
Citations
1,150
h-index
11
Badges

Publication numbers from OpenAlex, as of 18 October 2026.

Timeline

  1. 2021–2022

    From FastMoE to GLM-130B

    In 2021, he coauthors FastMoE, a PyTorch system for distributed mixture-of-experts training. In 2022, he coauthors GLM-130B, a bilingual language model with 130 billion parameters.

  2. 2023

    Tuning language models for agents

    In 2023, he coauthors AgentTuning, a method that enhances LLM agent abilities while maintaining general language-model capabilities.

  3. 2024

    Measuring long context

    He coauthors LongBench, a bilingual multitask benchmark for long-context understanding, in 2024.

  4. 2024

    Building a spoken chatbot

    In 2024, he coauthors GLM-4-Voice, an end-to-end chatbot for real-time spoken conversations in Chinese and English.

  5. July 2025–August 2025

    Vision and agentic GLM models

    In July 2025, he coauthors GLM-4.1V-Thinking and GLM-4.5V, vision-language models for multimodal reasoning. In August 2025, he coauthors GLM-4.5, an open-source model for agentic, reasoning, and coding tasks.

  6. August 2025

    Discussing smaller MoE models

    At GLM-4.5's public AMA in August 2025, he says the team will consider releasing smaller mixture-of-experts models.

  7. 2026

    GLM-5 and IF-CRITIC

    In 2026, he coauthors GLM-5, a foundation model aimed at agentic engineering, and IF-CRITIC, an LLM critic for instruction-following evaluation.

  8. Since 2024

    Z.ai researcher and Tsinghua PhD student

    He joins Z.ai as a researcher in 2024 and is a computer-science PhD student at Tsinghua University, where he starts in 2021.

Career

  1. Since 2024Zhipu AIResearcher

Education

  1. From 2021Tsinghua UniversityPhD

From the biography and OpenAlex affiliations (the years of papers that list the organization), so some years are approximate.

Most-cited works

  1. GLM-130B: An Open Bilingual Pre-trained Model2022 · 281 citations
  2. ChatGLM: A Family of Large Language Models from GLM-130B to GLM-4 All Tools2024 · 183 citationsWith Li Juanzi
  3. LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding2024 · Annual Meeting of the Association for Computational Linguistics (ACL) · 169 citationsWith Li Juanzi
  4. xTrimoPGLM: Unified 100B-Scale Pre-trained Transformer for Deciphering the Language of Protein2023 · 108 citations
  5. AgentBench: Evaluating LLMs as Agents2023 · 75 citations

From OpenAlex, as of 18 October 2026.

  1. The company share of AI researchers’ top papers rose from 3% to 32%Includes Zhipu AI. We estimate where leading AI researchers worked when their most-cited papers appeared. The company share rose from 3% in 1975–1979 to 32% in 2020–2024.
  2. Stanford University trained or employed 27 of the 111 AI major-prize winners, more than any other universityIncludes Tsinghua University. Stanford University trained or employed 27 of the 111 AI major-prize winners, more than any other university. We compare doctorate and employment links for major-prize and best-paper winners.
  3. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes Tsinghua University. Our estimates of where the people shaping AI studied and where their careers led, from education records in cited biographies.
  4. Google has the highest combined h-index among AI organizationsIncludes Tsinghua University. Our estimates of combined and median h-index at AI organizations, using current affiliations and OpenAlex records.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Aohan Zeng yet. Send us a free-licensed photo.

  1. FastMoE: A Fast Mixture-of-Expert Training System arxiv.org. Checked 18 October 2026.
  2. GLM-130B: An Open Bilingual Pre-trained Model arxiv.org. Checked 18 October 2026.
  3. AgentTuning: Enabling Generalized Agent Abilities for LLMs arxiv.org. Checked 18 October 2026.
  4. LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding aclanthology.org. Checked 18 October 2026.
  5. GLM-4-Voice: Towards Intelligent and Human-Like End-to-End Spoken Chatbot arxiv.org. Checked 18 October 2026.
  6. GLM-4.5V and GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning zhipuai.cn. Checked 18 October 2026.
  7. GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models arxiv.org. Checked 18 October 2026.
  8. 智谱 GLM-4.5 团队深夜爆料：上下文要扩、小模型在路上，还承诺尽快发新模型！ infoq.cn. Checked 18 October 2026.
  9. GLM-5: from Vibe Coding to Agentic Engineering arxiv.org. Checked 18 October 2026.
  10. IF-CRITIC: Towards a Fine-Grained LLM Critic for Instruction-Following Evaluation aclanthology.org. Checked 18 October 2026.
  11. Aohan Zeng alphaxiv.org. Checked 18 October 2026.