Joe Steinbring :laravel_bw::vuejs_bw:
-
I spent $2,500 pre-ordering an m5 pro Mac Mini with 64 GB or memory with hopes of running Qwen 3.8 Flash Next. It is a mixture-of-experts model with 125B total, 6B active but it sounds like 64 GB still isn't enough. Qwen 3.8 27B might be a good plan B. It's a dense 27 billion parameters model.
I stopped paying attention to this space for a little while and suddenly mixture-of-experts models are a thing.
#AI #LocalAI -
If I paid a bunch of money to see this fight, I think that I would be a little upset.
(Source: https://www.tiktok.com/t/ZTB1QPyfP/)