- cross-posted to:
- technology@midwest.social
You must log in or register to comment.
I am right at this moment letting a local chinese model (Qwen3.8 27b) update my build environment to mimic upstream build process changes.
Tedious secondary work which the tiny model does just perfectly (and more thorough than I myself would do it).
As the hardware is already available from other projects, that will literally just cost me some cents in electricity cost, compared to probably 20+€ if I would let some Anthropic model do the same work.
So I am really not surprised.
You just don’t need the latest frontier models for most stuff any more…The new Qwen 3.8 and Gemma 4 models are honestly very good to the point I’ve stopped using proprietary models for most stuff, by running these on a Mac mini. No need for a data center with the energy consumption of Manhattan….
