source: techcrunch ai: why the rise of open source ai isn’t hurting anthropic … yet
level: business
decagon ceo jesse zhang argues that open source and frontier ai models aren't direct competitors. instead, they represent different stages of an ai deployment lifecycle. companies often use expensive frontier models to test and prove new use cases. once those use cases mature, they switch to cheaper open source alternatives. this pattern means frontier labs like anthropic can maintain high spending even as open source token volumes rise.
data from vercel and openrouter supports this view. on vercel's ai gateway, deepseek leads in token volume with over a third of traffic, but anthropic still captures more than half of total spending. openrouter shows deepseek v4 flash processing 5.3 trillion tokens weekly, far more than anthropic's opus 4.8 at 2 trillion. however, opus 4.8 costs about 23 times more per token, so it likely earns more revenue. these figures suggest frontier models hold onto premium pricing despite open source growth.
the market for ai tasks is expanding quickly, which helps frontier labs. new, complex use cases keep emerging that require top models. even as older tasks move to open source, frontier models dominate early-stage discovery. this two-tier structure may become a stable feature of the ai economy. frontier providers retain the most valuable part of the market: high per-token prices. for now, open source success doesn't threaten their revenue.
why it matters: understanding this dynamic helps ai developers and businesses plan model choices and budgets, knowing frontier models remain essential for new, hard problems.
source: techcrunch ai: why the rise of open source ai isn’t hurting anthropic … yet