The rapid rise of open source AI is reshaping how companies build and deploy intelligent products, yet it has not significantly reduced demand for premium frontier models. A recent view from Decagon CEO Jesse Zhang suggests the two are not direct rivals, but rather parts of a broader adoption cycle.
In this framework, high-end models are often used first to test ideas, validate workflows, and unlock early value. As those use cases mature, companies increasingly shift to lighter and more affordable open source systems for production. At the same time, new and more demanding tasks continue to emerge, keeping overall spending on top-tier models resilient.
Usage data from platforms such as Vercel and OpenRouter reflects this split. Open source and lower-cost models are gaining momentum in token volume, with names like DeepSeek and Z.ai rising quickly. Yet premium providers still capture a large share of total spend, helped by their performance on complex, early-stage, and high-value workloads.
This pattern points to a two-layer AI economy: frontier models lead discovery, while open source increasingly supports scaled deployment. For businesses, that means more flexibility in choosing the right model for each stage of growth. For the industry, it suggests that innovation and efficiency may advance together rather than replace one another.
As AI adoption expands, this balance could define how the next generation of digital products is built and monetized.