Ring ring 🌴

Explore

Tips & Tricks 22
Microblog 21
Debug with Me 17
Uncategorized 9
Coding with AI 9
Soft Skills 3
Indie Journal 3
Life Story 2
Op-ed 2
Health 1

Local LLM benchmark: 7 model/host combos, 3 passes each, one agentic task

September 14, 2026

Seven local model/host combinations on an M5 Max with 128 GB, three passes each on the same agentic coding task, averaged: prefill, decode, memory, MTP acceptance, cache hit rate and…

murphitup @murphitup · 3d

This guy is so well spoken. I'm a big fan of the positive changes AI has made in my life, and optimistic about what good things the future could hold. But we need to move slowly enough that we're designing systems and incentives that protect _people_, not IPOs.
https://substack.com/@justindeschamps/note/c-326587767?r=2v9qcv&utm_medium=ios&utm_source=notes-share-action

Search for:

#microblog September 15, 2026 1 min read

1 Human

Authorship Info

This post was written by the following people, bots, and tools.

Human1

murphitup

Ring ring 🌴

Posted in: MicroblogTagged: Microblog

Local LLM benchmark: 7 model/host combos, 3 passes each, one agentic task

September 14, 2026

Seven local model/host combinations on an M5 Max with 128 GB, three passes each on the same agentic coding task, averaged: prefill, decode, memory, MTP…

That's all, folks!

Leave a Comment Cancel reply