China's Tech Platforms Build Custom AI Models to Cut Costs in Retail

China's tech firms are deploying in-house AI models to optimize retail services, with RedNote's recent release showing potential for cheaper goods but limited current impact.

China's technology companies are embedding custom AI models into consumer-facing platforms to improve operational efficiency, particularly in retail. RedNote, a lifestyle platform focused on fashion, travel, and shopping, released Dots3-Note Preview—a 280-billion-parameter open-weight AI model—on August 23, 2026. This model matched or outperformed major competitors like OpenAI, Anthropic, DeepSeek, and Z.ai in specific benchmark tasks. The development fits a broader trend where Chinese digital platforms bypass third-party AI providers to build tailored systems.

The mechanism lies in RedNote’s ability to create specialized AI that directly addresses retail workflows—like personalizing product recommendations or streamlining customer interactions—without relying on external services. This reduces costs and latency for businesses handling high-volume transactions. However, the current impact remains narrow: RedNote’s focus on lifestyle commerce means this shift applies only to specific retail segments, not the entire goods ecosystem.

For abundance in goods, this represents a potential pathway to cheaper, more responsive retail services through AI efficiency. Yet the effect is currently limited to China’s tech sector and specialized platforms. What to watch: Whether RedNote’s model scales beyond its niche to influence broader retail pricing or if the performance claims hold across real-world transactions. The source notes this is a narrower operational shift within China’s tech sector compared to wider AI adoption—meaning significant cost reductions for everyday goods remain unproven at scale. Current impact is constrained by the model’s specific use cases and the sector’s limited scope.

Source: SCMP Tech