The price of intelligence is collapsing. And that is not (only) good news
A task the best model solves for 6 dollars, another does for 72 cents and an open model for 13. What it means when intelligence becomes a cheap commodity, for those who use it and those who sell it.
Mattia PanattaUpdated 16 October 20266 minOpinion
Looking at Artificial Analysis rankings in October 2026, one figure strikes me more than the positions: how much it costs to have different models solve the same task.
The most capable model, Claude Opus 5.5, spends almost 6 dollars per benchmark task on average. GPT-6.1 Sol gets within a few points for 72 cents. The best open-weights model, Xiaomi’s MiMo-V2.6-Pro, is further behind but costs 13 cents. The first costs over 40 times the last. And every year, at equal quality, the figure drops again.
What changes for users
For individuals and small businesses this is great news. For under 10 euros a month you get tools that two years ago were reserved for big companies; through APIs, a website chatbot or an automation that reads emails costs a few euros a month in “intelligence”.
The practical consequence is that the cost of the tool is no longer the barrier. The barrier is knowing what to make it do, how to check it, how to fit it into your work. That is where the difference lies between those who benefit and those who don’t.
The less-told side
When something becomes cheap, much more of it gets consumed. It happened with energy, storage, photographs. With AI it means:
- More mediocre content. If writing a thousand articles costs a few cents, someone will write them. That’s the “AI slop” problem I discuss in another opinion.
- More total energy use. Cost per task falls, but tasks explode: agents working around the clock, research reading hundreds of pages. The overall data-centre bill rises.
- Thin margins for those selling “just AI”. If your product is an answer generated by a model, your price will follow the models’: downwards.
For those who work with clients
If you consult, write, design or teach, this dynamic affects you. The automatable part will lose market value, quickly. What stays valuable is what a cheap model can’t give: accountability for a choice, knowledge of the client’s context, trust.
That’s why, in my work, I price on outcomes and on the attention I give, not on the hours of “production” that AI has made almost free.
In one line
Artificial intelligence is becoming a cheap commodity; what stays expensive is knowing what to do with it and answering for it.
Sources
Informational content, current as of the date shown. Prices, models and rules change often: always check official sources. Written with the support of AI tools and reviewed by the author.
