Skip to content
TopicTracker
From HackerNewsView original
TranslationTranslation

GLM-5.2

ZAI has released GLM-5.2, an open-source multilingual language model supporting over 100 languages, available in 1B and 8B parameter sizes under the MIT License.

Background

- GLM-5.2 is an open-weight large language model released by the Beijing-based startup Zhipu AI (zai-org) on Hugging Face, as of March 2025 the strongest openly available Chinese LLM and reportedly on par with GPT-4o/Claude 3.5 Sonnet on certain benchmarks. - Zhipu AI is one of six Chinese "AI tigers" (alongside Baidu, Alibaba, Tencent, ByteDance, and Baichuan) and a spin-out from Tsinghua University's knowledge engineering group. It has raised over $1B from investors including Sequoia China, AliCloud, and Tencent. - The model is explicitly targeted at Chinese-language use and was trained partly on the now-restricted Chinese AI supercomputing cluster in Beijing, highlighting China's ongoing efforts to catch up with US foundation models despite US export controls on advanced chips. - This release is significant because it shows that Chinese AI labs can still train frontier-competitive models using domestically available hardware, though early independent evaluations caution that its English-language performance may lag behind top US models.

Related stories

  • The author tests the new open-weights GLM 5.2 model against Gemini 3 Flash by having them play text adventure games within a fixed budget of $0.15 per attempt. A harness tracks achievements earned in each game to compare the models' performance.