GPT-5.1, and what a point release tells you

OpenAI released GPT-5.1 on 12 November 2025, an update to the model that arrived in August. Not a new generation — a refinement of one three months old.

For most readers the release itself needs no action. The pattern behind it is worth thirty seconds of attention.

The cadence has changed

Between 2023 and 2025 the industry produced a new numbered generation roughly every nine to twelve months, each a visible step. What has arrived since August is a series of point releases: smaller, faster, cheaper, better at specific things.

That is what a technology looks like when it starts to mature. Not stagnation — consolidation. The same thing happened to smartphones around 2015 and to web browsers around 2012.

Why this is good news for a buyer

Because it makes planning possible. A market where a generational leap could arrive at any moment rewards waiting; a market of incremental improvement rewards deciding and getting on with it.

If you have been holding a decision because something much better might land, the shape of the last three months is the argument against continuing to wait. Something better will land. It will be ten per cent better, not twice as good.

The one practical consequence

Point releases change behaviour in small ways, and small changes are exactly the ones that go unnoticed until something downstream breaks.

If you have processes that run on a schedule, this is the version-pinning question from 13 August again. It has not become less relevant; it has become more routine, which is a different thing.

Two more significant launches are expected within the next fortnight. Tomorrow’s post covers how to evaluate one of them properly in a single afternoon.