Mike A

engineer | local llm

[Mike A @im_mike_a]

@uctive Awesome!!!
Thanks. We’re out hunting—will sit down and read this with him tonight. Saving while we have service!

[Mike A @im_mike_a]

@Marduk21 @murasametech Man that’s like an iPad model.
Strata will recommend a model. With that much ram it will be miles better than this.

[Mike A @im_mike_a]

@murasametech I ran strata on a 12 gb 5070 on a mini pc with 64 gb ddr5 and i9.

It ran Qwen 3.8 flash next IQ3_S at about 40-45 tok/sec.

[Mike A @im_mike_a]

@uctive Look forward to reading about it 👍

[Mike A @im_mike_a]

@uctive Sweet. I’m getting one for my kids is build then. This way they can buy it themselves and have some ownership. And tell their kids someday how primitive their hardware and ai was 😆

[Mike A @im_mike_a]

@uctive I know, and just so you know I actually meant it when I said it’s awesome. My 10 year old son and I are sitting here trying to figure out a machine he can earn and then build. You bet we’re looking at your ideas here 😎

[Mike A @im_mike_a]

@uctive Hahaha awesome

[Mike A @im_mike_a]

@DeForestIsMe @ryzerth Mind if I ask why you regret it? Is this for local ai machine?

[Mike A @im_mike_a]

@TeksEdge Just use Strata to get this performance for $3000 or less. I say this as someone who has experimented with good AI hardware (Dgx spark, 512 studio, rtx spark) , but is just building my first Strata build this weekend.

[Mike A @im_mike_a]

@coldniko @QwenDevs Sincerely hoping you get early access to the weights!

[Mike A @im_mike_a]

@Oluwaphilemon1 Good time to be ready. This will compound with Strata progressing and qwen4 models coming out.

[Qwen @Alibaba_Qwen]

The becoming begins.

[Mike A @im_mike_a]

@LINKPrepper @MiaAI_lab I grabbed this exact same deal. Good luck with yours. mine keeps stalling and restarting even without running a model.

[Mike A @im_mike_a]

@MattBinder You are trash.

[Mike A @im_mike_a]

@u1tra_instinct I got the same one thanks to your post! Thanks!

[Mike A @im_mike_a]

@gofish99983 @0xSero Me too. I have a mini pc I may try this with. But the Mac OS Linux thing is interesting

[Mike A @im_mike_a]

@0xSero think what this turns into as the models progress

[Mike A @im_mike_a]

@jasonm_sirois @stevenmarkryan @elonmusk The problem is that you can’t uninvent something, or even slow down advancement worldwide. It’s a prisoners dilemma. Even if all the American and Chinese AI labs completely stopped, it would just slightly delay progress and move the lead to different hands.

[Chad Hurley is Volatile Markets @volatilemarkts]

@NVIDIAAI … 300,000=DGX Station don’t you think??