This New Chinese AI Is Outperforming ChatGPT — and It Runs Locally
The 753-billion-parameter model uses an unrestricted MIT license and a one-million-token window, and Zhipu says it outperforms several US rivals on coding tests.
- Chinese AI startup Zhipu AI launched GLM-5.2 last week, an open-weight model featuring a 1-million-token context window for processing large codebases and agentic workflows.
- GLM-5.2 competes with frontier models like Anthropic's Claude Opus 4.8 and OpenAI's GPT-5.5, challenging gated, proprietary systems by giving developers direct access to the model's weights.
- On Monday, investor optimism drove Zhipu's valuation to HK$1 trillion , while the model scored 62.1 on SWE-bench Pro, surpassing GPT-5.5's 58.6.
- Developers are successfully running the 744-billion-parameter GLM-5.2 locally, though doing so requires aggressive compression to overcome the 'VRAM Wall' and hardware constraints.
- This shift suggests a future where companies combine flagship closed models for complex reasoning with specialized open-weight models for routine, private workflows at lower cost.
12 Articles
12 Articles
This new Chinese AI Is outperforming ChatGPT — and it runs locally
Developers are flocking to a new open weight AI model that can be downloaded, customized and run entirely on local hardware. Released by Z.ai (formerly Zhipu AI), GLM 5.2 is turning heads by challenging a major industry assumption: that you always have to pay premium subscription prices to a tech giant to get frontier-level performance. Unlike completely closed systems like ChatGPT or Claude, GLM 5.2 gives developers direct access to the model i…
What is GLM-5.2, China’s latest open-weight AI model turning heads in Silicon Valley?
Since its debut, Zhipu AI's GLM-5.2 has been generating a lot of buzz across social media with investors, founders, and developers expressing awe at the high-performance and capabilities of the new model.
Google Engineer Explains Why China's GLM 5.2 No Longer Needs To Distill American Models To Reach Mythos-Like Capabilities
It’s now widely accepted that China has been distilling American models to build its own, but it may no longer need to do... The post Google Engineer Explains Why China’s GLM 5.2 No Longer Needs To Distill American Models To Reach Mythos-Like Capabilities appeared first on OfficeChai.
Coverage Details
Bias Distribution
- 40% of the sources are Center, 40% of the sources lean Right
Factuality
To view factuality data please Upgrade to Premium
