---
title: "Discussion on \"Why did one way of splitting Meta's 816B model run 59% faster on the same two Macs?\" | Hashnode"
description: "Discussion on \"Why did one way of splitting Meta's 816B model run 59% faster on the same two Macs?\". Apple ran Muse Spark Open across Mac Studios. Split by layers, it decoded 15.74 tokens a second. Split inside each layer, 24.99. The reason is the parallelism trade-off interviewers ask about."
url: "https://hashnode.com/posts/why-did-one-way-of-splitting-metas-816b-model-run-59-faster-on-the-same-two-macs/6ad3493be282d6003117a7ed"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Why did one way of splitting Meta's 816B model run 59% faster on the same two Macs?](https://systemdesignacademy.hashnode.dev/why-did-one-way-of-splitting-metas-816b-model-run-59-faster-on-the-same-two-macs?utm_source=hashnode&utm_medium=feed)

Apple ran Muse Spark Open across Mac Studios. Split by layers, it decoded 15.74 tokens a second. Split inside each layer, 24.99. The reason is the parallelism trade-off interviewers ask about.

Oct 17 · 8 min read · systemdesignacademy.hashnode.dev  
Read the article

-   [#llm](https://hashnode.com/tag/llm)
-   [#machine-learning](https://hashnode.com/tag/machine-learning)
-   [#distributed-systems](https://hashnode.com/tag/distributed-systems)
-   [#apple](https://hashnode.com/tag/apple)
-   [#inference](https://hashnode.com/tag/inference)

## Written by

[RONI DAS](https://hashnode.com/@ronidas)

@ronidas · India

  
[View profile](https://hashnode.com/@ronidas)

AI engineering, agentic memory, machine learning, and system design at scale.

I work on AI systems and the infrastructure they actually run on: agentic architectures, memory for agents, retrieval, and the graph and data layers that keep model output grounded in something real. Formerly at Neo4j. I also write and teach the systems side that most AI content skips, from distributed systems and databases to the design decisions that decide whether a model ever reaches production. I run Total Technology Zonne on YouTube and I am building systemdesign.academy, an interactive course on system design, machine learning, and AI engineering, covering the parts most tutorials leave as an exercise for the reader. Mock interviews and prep packs: topmate.io/roni

## Responses

Markdown supported · ⌘↵ to post

No responses yet.