Open-source AI surges, but Anthropic keeps commanding premium prices

The tension between open-source and frontier AI models reflects a fundamental shift in how businesses deploy artificial intelligence. Rather than a simple winner-take-all competition, the market is developing into a two-phase lifecycle where expensive frontier models like Anthropic's Opus prove out new use cases, which then graduate to cheaper open-source alternatives as they mature. The data from Vercel and OpenRouter supports this picture: even though DeepSeek and other budget models now dominate raw token volume, Anthropic's premium pricing keeps it capturing the bulk of actual spending.

This arrangement may prove more durable than earlier predictions that foundation labs would become mere commodity suppliers. The fact that Anthropic's share of overall spending remains above 50% despite recent price increases suggests clients are willing to pay a premium for frontier model quality where it matters—particularly for novel, high-difficulty tasks. As more mature use cases shift to open source, new applications and harder problems continuously emerge to sustain demand for state-of-the-art models. The emergence of Nvidia's Nemotron as another competitor poised to take significant share does not yet challenge this basic pattern; instead, it adds another player to a market where frontier providers and open-source alternatives occupy different parts of the lifecycle.