Why the rise of open source AI isn’t hurting Anthropic … yet
TL;DR
Decagon CEO Jesse Zhang frames enterprise AI as a lifecycle: costly frontier models validate new use cases, while cheaper open source models later take over mature production workloads. Vercel data shows DeepSeek leading by token volume, while Anthropic still captures more than half of AI spend on the platform. OpenRouter points in the same direction: DeepSeek V4 Flash handles more tokens than Opus 4.8, but Opus costs about 23 times more per million tokens.
Nauti's Take
Anthropic has not been crushed by open source because the pricing power sits in a different column. Token volume can move quickly to cheaper models, while budgets stay with systems that make new or risky workflows work in the first place.
The risk comes later: once enough workflows are standardized and open source catches up on reliability, tool use, and operations, the premium markup gets harder to defend.
Briefingshow
For enterprises, the key question is not which model handles the most tokens, but where spend actually concentrates. If frontier labs dominate early workflow discovery while open source scales proven workloads, AI settles into a two-tier market: premium intelligence for uncertain tasks, cheap tokens for repeatable production.
Sources