For a couple of years there was a comfortable story about AI: a handful of famous American labs made the powerful models, and everyone else used them. That story just stopped being true, and the change is quietly good news for your business — if you understand which part of it actually matters.
The week the gap closed
On 8 July, xAI released Grok 4.5, claiming strong results on coding and knowledge work while using fewer tokens. Around it came a wave of others — DeepSeek V4, Qwen 3.6, GLM-5.2 — several of them open, meaning you can download the weights and run them yourself. NVIDIA even released an open model that pushes more than twice the throughput while keeping almost all of its quality.
The headline is not that any one of these “wins.” It is that the distance between the famous closed models and the free, open ones has shrunk to the point where, for a great many everyday tasks, you genuinely cannot tell the difference. The premium model and the free one give you the same answer to the same question.
Why a commodity is good news
When something powerful becomes abundant and cheap, that is not a threat — it is a gift, as long as you are not the one selling it. Electricity is a commodity; nobody frets that they cannot invent their own power station. They just plug in and get on with the work that only they can do.
AI models are becoming that plug. The cleverness you used to pay a premium for is turning into a utility you can pick off a shelf, swap out next month, and stop worrying about. That frees up your attention for the part that is not a commodity and never will be: your customers, your data, and the judgement about what to do with both.
Where your advantage actually lives
Two companies can use the exact same open model. One gets vague, generic results; the other gets sharp, useful ones. The difference is never the model — it is what each one feeds it. Your own clean, well-organised, consented data is the thing no competitor can download. It is the reason the same engine produces a better answer for you than for them.
So the right response to the open-model surge is almost boringly calm. Do not chase the model of the week. Keep your setup swappable so you can benefit from every new release without rebuilding anything. And put your real effort where it compounds — into the data and the rules underneath, the one layer that stays valuable no matter which model is winning the leaderboard this Tuesday.
The models racing to catch each other are doing you a favour. They are turning the expensive part of AI into the cheap part, and handing the advantage to whoever owns the data. That has always been you. It always will be.