Trending storyDeveloping
Zhipu and Tsinghua release open-weight GLM-4.6
1 report1 sourceupdated 2 hours ago
What happened
AI digest
Z.ai (Zhipu) and Tsinghua University released the open-weight model GLM-4.6 on September 30, according to an October 3 report. The report lists 3.6e11 parameters, 2.3e13 training data points and 120 days of training. It says GLM-4.6 scored 85% on Terminal-Bench 2.0 and 1820 on Text Arena (Coding), both above GPT-5, Claude Sonnet 4.5 and Grok 4.
Written by AI from the coverage · updated 2 hours ago
Coverage
Follow the reports to see the story from different sides.
Oct 3
- AI HOT · ModelsZ.ai 与清华大学发布开源权重模型 GLM-4.6
Z.ai（智谱）与清华大学于 9 月 30 日发布开源权重模型 GLM-4.6，参数规模 3.6e11，训练数据 2.3e13 条，训练耗时 120 天。该模型在 Terminal-Bench 2.0 上得分 85%，Text Arena（Coding）得分 1820，均高于 GPT-5、Claude Sonnet 4.5 和 Grok 4。
Heat over time
Not enough continuous observations to draw a trend yet.