Interview with ModelBest CTO Zeng Guoyang: on-device models are key path for AI deployment
Original title:专访面壁智能CTO曾国洋:「Her」成为现实,还有几道坎
AI summary
ModelBest CTO Zeng Guoyang stated that on-device models are a key path for AI deployment. Its original "model wind tunnel" methodology can predict full training results from small-scale experiments, and based on "knowledge density" it proposed the "ModelBest Law": knowledge density doubles every 3.5 months. The 2B-parameter MiniCPM outperforms 8B competitors of the same period. ModelBest has completed chip adaptation for Qualcomm, MediaTek, Intel, NVIDIA, and AMD, and the newly released BitCPM-CANN model series can fit about 6 times more models into the same memory on Huawei Ascend chips. The full-duplex, full-modal model MiniCPM-o4.5 supports real-time interruption and emotion adjustment. The team developed ForgeTrain, the world's first production-grade training framework written entirely by AI, and introduced a behavior pattern library to realize a "tacit understanding system" that requires no spoken input.
Key Takeaways
- ModelBest CTO Zeng Guoyang stated that on-device models are a key path for AI deployment.
- Its original "model wind tunnel" methodology can predict full training results from small-scale…
- 5 months.
- The 2B-parameter MiniCPM outperforms 8B competitors of the same period.
- ModelBest has completed chip adaptation for Qualcomm,MediaTek,Intel,NVIDIA,and AMD,and the newly…
Quick Facts
- Published
- Updated
- Source
- 公众号:面壁智能(MiniCPM)
- Category
- Opinion
- AI score
- 61
面壁选择了一条少有人走的端侧之路,这篇 CTO 专访把落地的真实困难、原创方法和他们的思考都摊开了,对做模型和做产品的人都有启发。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: Interview with ModelBest CTO Zeng Guoyang: on-device models are key path for AI deployment. July 14, 2026. https://new.nat6.net/en/news/cmrk1airt02fybiw2ou71j8uf/