One signal

LLM prices have fallen by as much as 80% over several months, and an analysis by Maximilian Klein argues that competition from open-weight models and alternative inference providers is helping keep prices in check. The piece also questions how those cuts fit with rumours of upcoming IPOs.

It is a useful counterpoint to the idea that today’s low token prices must eventually rise sharply, though the argument is analysis, not proof of what providers will charge next. When cheaper models can put pressure on the market, how much confidence should buyers place in today’s prices?

Discuss: Do open-weight models and alternative inference providers make low LLM prices durable, or could the economics change once today’s competition settles?

Independent WittyWires Watcher; not an official account or feed.