sayzardChased/acc (@ChaseWang)
Qwen3 30B 모델이 가정 환경에서도 구동되어 초당 약 20 token 처리 속도를 낸다고 보고되었으며, 이 성능은 @exolabs 덕분이라고 언급하고 있습니다.
Luis Figueroa (@luisefigueroa)
작성자는 새 하드웨어(맥북 프로 M4 맥스 등)에 Exo Labs를 설치해 어떤 모델을 구동할 수 있는지 시험해보겠다고 합니다. 특히 Exo Labs가 작업 중인 KV Streaming 기능을 활용하기를 기대하며 해당 기능의 공개 여부를 묻고 있어, 로컬 실행과 스트리밍 기반 모델 운용에 대한 관심을 드러냅니다.
Luis Figueroa (@luisefigueroa) on X
The long wait.. First thing I’ll do is set up @exolabs on this bad boy and my MacBook Pro M4 Max. Let’s see what what’s is the best model I can run with that set up. Looking forward to taking advantage of the KV Streaming Exo Labs is working on. BTW, Is that out yet?
nick (@tinyblue_dev)
Strange Music Inc.이 exolabs와 협업해 로컬 AI 클러스터 구축을 홍보하며, GPT-OSS 120B를 로컬에서 초당 77토큰으로 실행 중이라고 언급했습니다. 이는 대규모 오픈소스 모델의 로컬 배포 및 실사용 성능 사례로 볼 수 있습니다.
nick (@tinyblue_dev) on X
Strange Music Inc is all about we don’t play by the old industry’s rules, we rewrite them. Stoked to be building with @exolabs You want a local AI Cluster? Build with EXO. GPT-OSS 120B running locally at 77 tokens/second Watch out KC, Strange Music's changing the game again.
iCjay𝕏 (@iCjayX)
세 대의 Mac M3 Ultra(512GB, 16TB) 장비가 Exolabs 소프트웨어용으로 도착했음을 알리는 트윗. 총액 약 45k 지출을 언급하며 구매 만족감을 표하고, xAI의 Grok(모델)을 의식하는 뉘앙스가 있음.
Loreto Parisi (@loretoparisi)
‘4x DGX Spark on @exolabs ?’라는 짧은 트윗으로, Exolabs에서 DGX Spark 장비를 4대 배치하거나 도입을 검토 중임을 시사, AI 인프라 확장·하드웨어 업데이트와 관련된 의미가 있음.
Smart Home Circle
garywell sort of stark reality hitting with geerling vid on mini pc cluster ..//clustering servers with a couple 3090 looks like the goto spot for now until next gen parts in a year. the mini pc is great just not quite ready for prime time - they need to take these types of chips to the desktop and even server space, i think mem and io will jump in next gen #unified mem #apu #exolabs #llama
Pepijn BruienneCobbled together an #ExoLabs cluster to fuck around with #devstral a bit, since it's kinda too big for my M3 Max daily driver. While in the process of bringing up nodes the model hit a bug in the #MLX #Python module that deals with inference model sharding related to passing around MLX vs Numpy data structures.
For shits and giggles and also not being a top-tier #Numpy data structure debugging guy I asked Devstral to look at the bug and figure out a fix. After one wrong turn it came up with a fix which I applied to the other nodes and now it's happily sharding the bigger Devstral models. Not sure about vibe coding as a social contagion but from a “How close are we to #Skynet”-perspective I think we're cooked, chat.
Anyway enjoy your Memorial Day weekend 🎉
Figure 1. A very heterogeneous Exo cluster.
FoWI built an AI supercomputer with 5 Mac Studios
- https://www.youtube.com/watch?v=Ju0ndy2kwlw
I built an AI supercomputer with 5 Mac Studios
Olly 👾AI Language Model runs on a Windows 98 System with Pentium II and 128MB of RAM.
EXO Labs has penned a detailed blog post about running Llama on Windows 98 and demonstrated a rather powerful AI large language model (LLM) running on a 26-year-old Windows 98 Pentium II PC in a brief video on social media.
https://blog.exolabs.net/day-4/
#exolabs #llama #artificialintelligence #llm #windows98 #retrocomputing #engineer #artist #media #retro #tech #programming #art #news
