The cost argument is wrong. Compare Gemini 2.5 Pro to Gemma 4 31B.

Released only 10 months later, a tiny open-weights model outperforms what was once SOTA. Fable and 5.6 Sol will be outperformed by laptop-class models next year.

If you only ever want frontier model performance then sure, you have to pay to play. But as it is now with open models, some of which I can even run from my gaming PC at home, we're only about 6 months behind frontier performance. Even when the money starts drying up for those at the forefront, the genie is out of the bottle.

And that's why I pivoted at the beginning of this year to LocalLLMs.

I can afford the hardware I already have. And I can run jobs on it as I see fit. Sure, its slower than the corpo LLMs, but we're already seeing lying, silent downgrading, and advertisements, even when paying for tokens.

When your systems goes tits up or priced out, mine will still work.