A major cloud platform closed a catalogue of 11,000 models available to corporate customers in the period, including models from direct competitors.
The number describes a change of problem. Two years ago the difficulty was getting access to a capable model; now it's choosing among thousands, with no established comparison criteria for specific uses.
A large catalogue favours whoever operates the platform, not necessarily whoever uses it. For the customer, more options mean more evaluation work, and the natural tendency is to settle on the most-cited model, which is rarely the best fit for the task.
Competitors' models sitting inside the catalogue confirm what other products in the period already showed: models became an input, and a platform reselling inputs wins by offering variety rather than defending its own.
For anyone choosing, the criterion that survives any catalogue remains measuring on your own workload, with tasks your operation actually runs, comparing cost per completed task rather than price per token.