Anonymous Niu Lai Large Model Found to Have Zhipu Lineage, Some Also Suspect Cursor Trained on Open-Source GLM
Community analysis indicates an anonymous large model is highly similar to Zhipu GLM in aspects such as Tokenizer and API error messages, raising doubts about whether Cursor was trained based on open-source GLM.
Community tech enthusiasts conducted deep reverse analysis on an anonymous large model, finding high overlap with Zhipu's open-source GLM series in Tokenizer tokenization logic, video encoding methods, and API error messages.
This similarity in technical fingerprints has sparked industry attention regarding model lineage. Although open-source models allow commercial use, directly reusing core components without significant improvements may involve intellectual property boundary issues.
The discussion also extended to the AI coding tool Cursor, with opinions suspecting its underlying model might be fine-tuned based on open-source GLM. This reflects the current status quo where open-source weights are widely utilized under large model competition.
Currently, related accusations remain at the technical analysis level, with no official response yet. This incident reminds developers to clarify authorization scopes when using open-source models, while also demonstrating the community's improved ability to trace model origins.
This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.