️ GLM-5: China's New Best AI Model
️ GLM-5: China's New Best AI Model
The new GLM-5 AI model from Z.ai has become the smartest open weights model. According to Artificial Analysis, GLM-5 reaches the Claude Opus 4.5 level while costing about 5-7 times less. GLM-5 also ranks above Gemini 3 Pro and GPT-5.2 Codex.
Key Features:
Almost no hallucinations. Right now, GLM-5 is the best model with a 66% score on the AA-Omniscience benchmark, which tests AI's tendency to make up details. Claude Sonnet 4.5 comes in second with 52%.
Great at coding. In programming and agentic tests, GLM-5 performs as well as closed-source leaders, and even beats them in some cases. However, the model still noticeably trails GPT and Claude on scientific tasks.
Despite being open weights, you won't be able to run GLM-5 on home hardware. A model with 744B parameters (40B active) requires at least several hundred gigabytes of VRAM.
️ Still, you can use GLM-5 for free on the Z.ai website.
Source: Telegram "chinadream_EN"