Gavin Baker highlights Anthropic’s lower cost per token than OpenAI

In the race to dominate AI, the most important metric might not be who builds the smartest model. It might be who builds the cheapest one. Gavin Baker, managing partner and CIO of Atreides Management, has been making a pointed case that Anthropic holds a decisive edge over OpenAI in one of the most consequential dimensions of AI competition: cost per token. Speaking on the Invest Like the Best podcast, Baker laid out a thesis that reframes how investors should evaluate AI labs, and the numbers are striking. The efficiency gap Baker’s core claim is that Anthropic burned roughly 80% less capital than OpenAI to reach a similar revenue run rate. The source of this efficiency advantage, according to Baker, is twofold. Anthropic’s models achieve higher “intelligence density per token,” meaning they extract more useful output from each unit of computation. And they simply use fewer tokens to get the job done. Claude Opus, Anthropic’s flagship model, reportedly generates around 70% fewer tokens per query compared to earlier iterations. In practical terms, Anthropic can deliver comparable answers to competitors while consuming roughly half as many tokens. Baker frames this through what he c...