Vercel's data suggests a pivotal shift is underway, moving beyond simple cost-cutting to a fundamental re-sorting of the model stack. The key development isn't just DeepSeek's volume growth; it's that newer open-weight models like Kimi K3 and GLM 5.2 are, according to the report, beginning to capture spend on workloads previously reserved for closed-weight labs. This implies enterprises are starting to trust open models for more complex, agentic tasks, not just bulk processing.
If this trend holds, the competitive moat for frontier labs may be shallower than assumed, hinging less on raw capability and more on ecosystem lock-in and reliability. The industry should watch for whether this spend shift accelerates or if closed labs can counter with new pricing or bundling strategies.