{"type":"rich","version":"1.0","author_name":"npub1rr9nrkygq565nv6hu3e25ypjlvgjnwnxmwcp7pv7u9zsr0kwtn7sjdpfl2","author_url":"https://nostr.ae/npub1rr9nrkygq565nv6hu3e25ypjlvgjnwnxmwcp7pv7u9zsr0kwtn7sjdpfl2","provider_name":"njump","provider_url":"https://nostr.ae","html":"nostr:npub1y6ckppt03huzdsaz472u5ue7yf3v5wp5y3ygvpf74mlthee54sks2xjkf0 \u003e What do you expect to get out of local AI?\n\nI don't.  I was shocked that anything ran on my machine at all, since it only has Intel onboard graphics.  I was just curious enough to try it and see if it worked.\n\nI mean we're only talking 3-10 tokens/sec, so it's not like it's remotely fast, but I found it at least usable.\n\nThe machine Rob Braxman recommends is this one:\n\nhttps://www.amazon.com/Beelink-126TOPS-LPDDR5X-PCIe4-0...\n\n\u003e Basic image,music, and chat can do great on 12-16GB GPU\n\nI'd expect less than that for images if the Ollama ever gets the flux2-klein model working on Linux.  I haven't even look at music generation though."}