As much as I am excited about cool LM uses, right now they are pretty much interchangeable and a commodity. I doubt many people could pass a blind taste test.
How this goes:
- new model released
- bullshit synthetic benchmarks
- SVG pelican with subjectively different qualities than previous drawings
- accusations of pelicanmaxxing
- the exact same discussion: should I pay a lot of money to run locally, or should I pay a lot of money to a tech company who is absolutely mining my data?
- user claims to have 10,000 hours on it and it’s shit
- user claims to have 30 seconds on it and has switched over their entire life to it and will never need another model so long as they live
- new model released