If you're wondering why Gemini limits have plummeted dramatically, the answer is simple, they are selling compute to other companies, google realized they can make MUCH more money per token selling it to anthropic than using their own models
Instead of investing into strong models, they started releasing countless fluff empty products to mask this transition.
Which explains the "flash" model too.
Google, at least for now, decided to turn into a compute supplier rather than a model maker it seems.
If you were stupid like me and bought for the whole year using the new year offer, you have been RUGGED.
In retrospect the deal was too good to be true, which means they were planning on doing this since then.
Which is why all their pre-train (base models) are STILL working on the 2024 dataset (jan 2025 cutoff).
AND another thing, they invested in anthropic as a way to position themselves to potentially acquire it in the future.
This is more of a conspiratorial thing but it makes sense, they have showed alot of interest in anthropic.