.jpg?w=1024)
The most powerful AI model is not always the most useful one.
Businesses are increasingly testing lower-cost Chinese models for coding, customer support, and internal automation, according to reporting from NPR and usage data from OpenRouter. Although leading US systems may still outperform them on advanced tasks, the price difference is prompting companies to reconsider how much intelligence they actually need for routine workloads.
That calculation could reshape the AI market. Chinese developers may not need to erase the performance gap if their models are reliable enough to handle high-volume work at a fraction of the cost.
Businesses choose practical AI over premium
In a conversation with NPR, Eugene Cheah, CEO of AI platform Featherless, compared the difference between expensive frontier models and cheaper alternatives to choosing between a luxury car and a reliable everyday vehicle.
Many companies, he said, do not need the “Ferrari” of AI for every task; instead, they are choosing lower-cost models that can handle large, routine workloads at scale.
Data from OpenRouter, a platform that provides access to multiple AI models, shows that Chinese AI models have gained significant adoption among developers in 2026. DeepSeek’s share of usage has increased, while models from companies such as MiniMax, Xiaomi and Tencent have also seen growing interest.
Many businesses are using these lower-cost models for repetitive workloads, including coding assistance, customer support and internal automation.
Large companies also explore Chinese AI
The shift toward Chinese AI models is not limited to startups, as larger companies are also experimenting with cheaper systems as AI expenses continue to grow.
Reports indicate that companies, including Airbnb and DoorDash, have used Chinese-developed models for certain workloads. Airbnb previously used Alibaba’s Qwen model, while DoorDash executives have highlighted savings from using lower-cost AI systems for less complex tasks.
Cost is not the only factor driving adoption. Open-weight models also give companies greater control over how AI systems operate, allowing them to customize the technology and better understand how data is processed. That control is becoming increasingly valuable as businesses consider data privacy, security and long-term dependence on AI providers.
Some still prefer US models
Despite the growing interest in Chinese AI, some companies remain cautious and committed to US-developed systems.
Jon Gordner, CEO of coding startup Comment.io, told NPR that his company continues using Anthropic and OpenAI models because accuracy and development speed matter more than short-term savings.
After all, for companies building complex products, cheaper AI models may create additional costs if employees need to spend more time correcting mistakes or improving results.
The choice depends on the task. Premium systems may remain valuable for advanced reasoning and research, while cheaper models are better suited for routine business operations.
A hybrid future for AI
Many experts believe companies will increasingly use multiple AI models instead of relying on a single provider.
Companies are increasingly adopting “model routing,” sending complex tasks to powerful AI systems while assigning simpler work to cheaper alternatives.
The rise of Chinese AI models is also putting pressure on US companies to respond through lower prices, improved efficiency and more flexible offerings.
As businesses become more cautious about AI spending, future competition may not simply be about which company builds the most powerful model. It may be about delivering reliable AI that companies can afford to use at scale.
Related reading: For another example of how Chinese AI developers are competing through efficiency and openness, see our breakdown of Kimi K3 and its open-weight approach.