Replies (29)

True. People are buying them with the wrong specs / intentions. A $2600 Mac mini will actually run decent Qwen models on board and can mitigate token costs with a good routing policy, and/or an operator who knows when to delegate to cloud models. What seems expensive up front can actually be a good investment saving hundreds in tokens per month. (Depends on user of course… you are probably right for most people)
My use cases don't justify running local models as they can't compete with Claude Opus. When that changes I'll run them locally, but not on a Mac.
Doing like u , laptop 2013. But the token costs are pilling for the basic stuff. I think in the end (6 months) a Mac mini running local LLm for majority of stuff will cost the same as my old laptop running all on pay per use tokens. Also some anonymity advantages. 100 usd here and there on tokens to configure files pile up quickly
You’re welcome. I plugged mine into an internal chat tool I have for my company so my staff and I can all dm or engage in the team chat. He’s nice and fast there. I also just ssh into his box a lot and interact from the shell. I hope to look into marmot protocol this weekend for more secure mobile chat than my company app
curt finch 's avatar
curt finch 2 months ago
Have you got it to do anything useful? PS if I'm the boss then when you do take over the world it's my world
Otto Kalvo's avatar
Otto Kalvo 2 months ago
Anyone with experience running it on the new Umbrel ?