Post

Conversation

M4 Mac Mini AI Cluster Uses

with Thunderbolt 5 interconnect (80Gbps) to run LLMs distributed across 4 M4 Pro Mac Minis. The cluster is small (iPhone for reference). It’s running Nemotron 70B at 8 tok/sec and scales to Llama 405B (benchmarks soon).

0:18

Can't wait for benchmarks + a write up! Performance overall, cost performance, thoughts on usability as an "AI Brain" for business and personal use cases, etc...

91K

Alex Cheema - e/acc

@alexocheema

Will be publishing a lot of benchmarks, not just for this setup but many different variations. All tested for real with real devices. Sneak peak:

5:32 AM · Nov 9, 2024

352.7K

Views

Post your reply

Matt McDonagh

@McDonaghMatthew

Nov 9

Dude that is BRILLIANT I love the layout, clean and easy to digest. Looking forward. We have so much cognitive leverage at our fingertips now.

exo x

are a match made in heaven i think, especially w/ INTELLECT-1

INTELLECT–1: Launching the First Decentralized Training of a 10B Parameter Model

From primeintellect.ai

Hit that follow button & be part of our growing #MobileDev community

We're your source for daily #Android & #Flutter blogs, job postings, #codelabs, & global events where thousands of #AndroidDevs & #FlutterDevs converge to inspire & get inspired.

Slide 1 of 4 - Carousel

Stay Updated on the Latest Mobile Dev Trends!

913K

Arthur Collé Organization

@arthurcolle

Nov 9

Wow the new UI is crazy, beautiful work!

damn this looks like it’s gonna be so useful

Omg amazing. Wen

Özv. Dízelné Hadházy Aranka, 1.8T

What is this UI?

But will it blend?

looking forward to it!

This is a custom UI. Ignore the names and numbers. It’s mock data to represent initial progress on the UI while benchmarks are being gathered. This was a sneak peak. The UI has been updated further. You’ll see real data and be able to use the site after benchmarks are ready.

We’re thrilled that

has selected

to bring the Bitcoin Lightning Network and Universal Money Addresses via the

@umastandard

to its platform and 100 million customers. We’re working with the Nu team on the technical integration and will have more to share in the

please do

3x studio display? Is this rage bait?

Just for your information, $20.000 for three machines, for running a 70b q4 model, doesn't make ANY F SENSE. I do the same in a little box with 4xA4000, total cost of the machine, less than $3000 You can even more with a mining rig and 6 or 8 cheap 4060ti 16GB ($450/each)

What is this UI?

Woo expecting to read the report

Kurwwwwwwwwa

@BardzoKurwa

Nov 10

1) This is interesting...the best board I could find is the ROMED8-2T and that only supports 7 GPUS. Let's say I get the highest spec'ed Mac Mini + cpu. That's 64GB. Now let's say I get 4 of them. That's 256GB RAM. $8,076.00 (with my edu discount at Apple). cont...

I love the intersection you guys are working in. Software, hardware, research and product. All the best, will keep supporting and wish to build something like this someday.