Free AI Is Catching Up to the Paid Stuff — Fast

A widely watched scoreboard called Artificial Analysis just posted a new number: the leading paid AI model, Opus 5.5, now sits at 58 points. The best "open" model — meaning software anyone can download, run, or modify for free — is 12 points behind. That's a big gap by scoreboard standards, roughly the difference between the very top tier and the tier just below it.

Why should you care about a scoreboard? Because open models are the free version of AI. You don't pay a monthly subscription, you don't send your data to a company's servers, and businesses can customize them however they like. For years, open models trailed the paid ones badly. That gap is now measured in months, not years. The post's author does the math: if free models gained 2 points in two months, they'd hit 58 in about six months — and possibly in four or five, since teams keep buying more chips and training on more data.

The catch is that these scores measure specific tasks, mostly coding and tricky reasoning puzzles. They don't measure whether an AI is pleasant to talk to, accurate about your documents, or safe to trust with real work. A model can score well and still be frustrating. Predictions like "four to six months" are also guesswork — progress in AI has stalled before, and running a big open model at home still requires expensive hardware or a paid cloud account.

So don't cancel your subscription yet. But do expect the price of good AI to keep falling, the way smartphone cameras made cheap cameras pointless. When free catches up to paid, companies usually respond by cutting prices, adding features, or both. The practical takeaway: powerful AI is on track to become a commodity, and commodities get cheap.