ModelBest open-sources MiniCPM5-1B: surpasses all sub-2B models on AA-Index and runs on phones and browsers
AI summary
ModelBest has open-sourced MiniCPM5-1B, its new-generation on-device large language model. With only 1B parameters, it surpasses all sub-2B models on the AA-Index leaderboard, outperforming Qwen3.5-2B from 3 months ago with half the parameter count. After INT4 quantization, the weights are only 0.5GB, supporting operation on phones and browsers. Its Base Model version was pretrained with ForgeTrain, ModelBest's self-developed AI training framework, and the model weights, training datasets and deployment solutions are now fully open-sourced.
Key Takeaways
- ModelBest has open-sourced MiniCPM5-1B,its new-generation on-device large language model.
- With only 1B parameters,it surpasses all sub-2B models on the AA-Index leaderboard,outperforming…
- 5-2B from 3 months ago with half the parameter count.
- After INT4 quantization,the weights are only 0.
- 5GB,supporting operation on phones and browsers.
Quick Facts
- Published
- Updated
- Source
- IT之家·人工智能
- Category
- Model Releases
- AI score
- 76
1B参数干翻所有2B以下模型,量化后0.5GB就能塞进手机和浏览器,这个效率把端侧模型的门槛又压低了,做移动端AI的值得跟进。
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: ModelBest open-sources MiniCPM5-1B: surpasses all sub-2B models on AA-Index and runs on phones and browsers. May 26, 2026. https://new.nat6.net/en/news/cmpm8mi1b0miusl01tr5zuge8/