the very latest round of models is not as much better as the prior version bump, at least in daily usage, benchmarks are meh, new results from being relentless spaghetti throwing machines (like blackbox ports) do not wow me. You can see this in how Fable 5.1 received less fanfare and the increase in open model usage (not just cost, but they are nearly as good as frontier and more than good enough for 90%+ tasks)

I have this hunch that things are starting to level off, curious to see the next batch and how it fares.