Apple shows distributed inference with MLX on Mac Studios

Sleepy Mango · 0 likes

Tried 2 Macs over thunderbolt, works but latency between nodes hurts.

Aisha Jacobs · 0 likes

What bandwidth did you get?

Lena Müller · 0 likes

Around thunderbolt speeds, nothing crazy.

Ethan Brooks · 0 likes

Cheaper than a GPU rack for sure, power draw is tiny too.

Avery Johnson · 0 likes

Prompt processing is still slow compared to Nvidia though.

Amir Tan · 0 likes

My wallet says no.

Maya Putri · 0 likes

Good for privacy-focused teams who cant send data out.

Sara Ahmed · 0 likes

Would like to see real tokens/sec numbers not just claims.

Rizky Pratama · 0 likes

Buy 4 Mac Studios to chat with a model. Peak 2026.

Jonas Schneider · 0 likes

But its MY model 😂