RedNote, the Chinese social media giant also known as Xiaohongshu, has made a significant move in the AI landscape by launching its open-source large language model, dots.llm1. This release marks RedNote’s entry into the open-source AI trend that many Chinese tech firms are embracing, diverging from Western proprietary models.
The dots.llm1 model, developed by RedNote’s Humane Intelligence Lab, boasts an impressive architecture that activates 14 billion out of 142 billion parameters, striking a balance between performance and cost-efficiency. Surpassing 11.2 trillion high-quality tokens during pretraining without synthetic data, the model’s capabilities are showcased on platforms like Hugging Face.
In contrast to Western tech giants like OpenAI and Google, which tend to keep their best AI models under wraps, Chinese companies like RedNote are opting for open-source strategies. This strategic divergence goes beyond business models, as it reflects a broader shift in trust frameworks that will shape the future of enterprise AI procurement.
RedNote’s dots.llm1 model, although not topping the charts like DeepSeek-V3, demonstrates solid performance, scoring 56.7 on the C-SimpleQA test for Chinese language skills. However, some analysts suggest that RedNote could leverage its data-centric strengths in content-driven commerce to develop more targeted AI models for enhanced business outcomes.
The economics of free AI are at play with RedNote’s dots.LLM1, designed with a “mixture of experts” approach that optimizes task-specific activations, reducing operational costs. This model serves as a market accelerant rather than a revenue product, positioning RedNote to establish foundational infrastructure and drive platform entrenchment through widespread adoption.
Moreover, Chinese firms benefit from governmental support, enabling sustainable strategies that might be financially challenging in Western contexts. By exporting open-source LLMs as soft power tools, Chinese vendors like RedNote extend their influence globally, shaping markets in regions like Southeast Asia, Africa, and Latin America.
RedNote’s foray into AI extends beyond model releases, with practical applications like Diandian, a search tool on its social commerce platform. With a substantial user base and recent international expansion, RedNote is positioning itself as a global player in AI, leveraging its AI capabilities to enhance user experiences and drive growth.
As enterprises navigate the landscape of open-source AI alternatives, considerations extend beyond cost and performance to encompass transparency, control, governance, and geopolitical implications. While the rise of free AI models presents attractive opportunities, organizations must carefully evaluate the trade-offs and implications for their AI strategies in a rapidly evolving technological landscape.