Tech companies switch to cheaper open AI models

Several tech companies are dropping the habit of using the most expensive AI model for everything. Instead, they're bringing in cheaper, open models for the tasks that don't require the big brain.

Not every task needs champagne and caviar; some do just fine with rye bread. That's why more companies are investing in "model routing": a kind of traffic controller that automatically sends simple requests to cheap models and only lets the hard tasks go on to the expensive ones.

At the same time, more companies are starting to set concrete caps on how much AI each individual developer may use. And knowledge sharing has taken off in the industry about how to get the most AI for the money.

For you as an individual, this doesn't change anything here and now: it happens behind the scenes at the providers.

If you run a company, it's worth reacting if your AI usage has started costing more than expected. Look at whether you're using expensive models for tasks that are actually fairly simple: that's typically where the savings are hiding.

What it means for you

If your company has its AI spending under control, nothing changes for you right now. But if you use AI a lot, you can save money yourselves by checking whether simple tasks are being sent to expensive models when cheaper ones could do the same job. Talk to your IT department about whether there are options for setting up similar rules for your own usage. As an individual, you won't notice it: it happens behind the scenes at the providers.