AI is facing its first price war, with users increasingly looking for value as well as capability. Frontier US labs such as OpenAI and Anthropic are months ahead of open-source rivals such as China’s DeepSeek. But advanced functionality comes at a price: Anthropic’s new Fable model is roughly 50 times as expensive per token as DeepSeek’s V4; it can do more complex things, but most customers “don’t need a model that knows quantum gravity,” one academic told The Wall Street Journal.

Indeed, OpenAI is mulling slashing prices, the Journal reported, while some customers are only using highly capable models for complex tasks, relying on cheaper open-source ones for most functions, saving up to 95% on outlays.