BEIJING, June 9 (Reuters) – China's Renote, one of the country's most popular social media platforms, is taking part in a wave of Chinese tech companies releasing large-scale open-source language models and making artificial intelligence models freely available.
This approach has also been released open source models for some American companies, including Meta, but in contrast to many US tech giants like Openai and Google.
Open sourcing allows Chinese companies to demonstrate their technical capabilities, build a developer community, and spread their influence worldwide when the US attempts to thwart China's technological advances with sophisticated semiconductor export restrictions.
A Rednote model called Dots.llm1 can be downloaded by embracing the developer platform. Company technical paper uploaded on Friday.
In task coding, the model runs relatively on Alibaba's QWen 2.5 series, but drags more advanced models such as DeepSeek-V3.
Renote, also known as the Chinese name Xiaohongshu, is a platform like Instagram where users share photos, videos, text posts and live streams. The platform attracted international attention when some US users gathered for the app earlier this year amid a potential concern about the ban on Tiktok.
The company has been investing in large-scale language model development since 2023, shortly after Openai released CHATGPT in the second half of 2023.
In recent months, we have accelerated our AI efforts to launch Diandian, an AI-powered search application that helps users find content on Xiaohongshu's main platform.
Other companies pursuing an open source approach include Alibaba, which launched its Qwen 3, an upgraded version of the model in April.
Earlier this year, Startup Deepseek released a low-cost R1 model as open source software.
(Reporting by Liam Mo and Brenda Goh, edited by Edwina Gibbs)
