Companies rethink tokenmaxxing as AI token costs fail to deliver
Companies have been all in on "tokenmaxxing," using as many AI tokens as possible to boost productivity. But now, the high price of these tokens isn't paying off, and businesses are starting to feel the pinch.
Tokens, which are basically tiny chunks of text AI uses to process information, come with steep costs that just aren't matching up with results.
Companies pivot to model routing
While big names like OpenAI's Sam Altman and NVIDIA's Jensen Huang once championed tokenmaxxing, companies are rethinking things.
Many are switching to "model routing," sending easy tasks to cheaper AI models and saving advanced ones for tougher jobs.
Open-source options from Chinese firms like Moonshot and Zhipu are also catching on for being both capable and budget-friendly.
As Mozilla's chief technology officer, Raffi Krikorian, put it, tokenmaxxing could soon be a thing of the past, kind of like counting lines of code used to be.