Chinese social media platform RedNote, also known as Xiaohongshu, has announced the open-sourcing of its first large language model (LLM), dots.llm1.

The Shanghai-based company unveiled the model on June 6, 2025. It features a mixture-of-experts system to activate 14 billion parameters out of a total of 142 billion in response to queries.

Developed by RedNote’s Humane Intelligence Lab, the model is designed to balance performance with lower training and inference costs. This lab evolved from the company’s former AI research division.

RedNote said that dots.llm1 surpasses other leading open-source models in understanding the Chinese language. This includes Alibaba’s Qwen2.5-72B-Instruct and DeepSeek-V3.

The company has 300 million monthly active users and is valued at about US$26 billion, according to Bloomberg.

It has also expanded internationally by opening its first office outside mainland China in Hong Kong on June 7.

🔗 Source: South China Morning Post

🧠 Food for thought

1️⃣ Open-source AI is becoming an economic strategy, not just a technical choice

RedNote’s decision to open-source its LLM aligns with a broader industry shift where economic factors increasingly drive open-source AI adoption.

Research shows that 89% of organizations using AI now leverage open-source models, with two-thirds finding them cheaper to deploy than proprietary alternatives 1.

Organizations implementing open-source AI report potential cost reductions of up to 50% in business unit operations, making it an attractive option for companies looking to maximize AI investments while controlling expenses 2.

This economic advantage explains why well-capitalized Chinese tech giants like RedNote, Alibaba, Tencent, and ByteDance are pursuing open-source AI despite having the resources for proprietary development. They’re responding to market demand for more accessible, cost-effective AI solutions.

The open-source approach also allows these companies to benefit from community contributions and improvements, effectively distributing R&D costs while maintaining competitive advantages in other areas of their business.

2️⃣ Strategic transparency addresses AI sovereignty and trust concerns

RedNote’s decision to release model checkpoints for every trillion tokens trained demonstrates a strategic approach to transparency in AI development.