It's astonishing how much community goodwill and excitement there was for llama when it first came out, and how much it has plummeted since then.
Partly the sheen has worn off, and of course there are other interesting "open source" llm offerings now, but I don't think it can be overstated how much Meta is just a toxic brand.
fwiw this really doesn't mesh with my experience. The goodwill and excitement is still there for Llama, though diluted by the abundance of model options out there. That said, my lense does tilt heavily towards self hosting where the community is probably overall much more appreciative and excited about Meta's AI
WRT Llama, IMO it’s only because they couldn’t infect a model with ad tracking that it overcame the tarnish. As long as I don’t need to run their conventional software to use it, I consider it safe.
I’m on the fence about their API product( technically they could tie it to one’s identity (as OpenAI could) but I’d imagine that would be less valuable given they can’t know which human is really behind it.
Partly the sheen has worn off, and of course there are other interesting "open source" llm offerings now, but I don't think it can be overstated how much Meta is just a toxic brand.
Threads had a very similar arc.