Дружба Linux и Windows, или как поиграться с ИИ-моделями на втором компьютере без видеокарты
Работаете под Windows? Есть лишний компьютер? Нет видеокарты, но хотите сами запускать ИИ-модели? Тогда делаем как на картинке, по шагам.
https://habr.com/ru/articles/1024950/
#Локальный_ИИ #Windows #LMStudio #openwebui #linux_mint #паранойя_mode_on
⚠️ Online-Dienste wie #ChatGPT, #Copilot & Co. benötigen immer eine Internetverbindung, weil sie Daten an fremde Server übertragen.
Dank der kostenlosen Software "LM Studio" kannst Du viele KI-Modelle direkt auf Deinem PC betreiben, also lokal und offline – auch von #Microsoft.
👉🏻 Meine neue Anleitung richtet sich an Einsteiger. Du kannst #LMStudio leicht selbst installieren und dann sofort nutzen: https://www.malter365.de/apps/lm-studio/
Some answers lack citations despite referencing empirical claims (e.g., economic system comparisons).
Read the full article: Assessment of Qwen3.5-9b in LMStudio
▸ https://lttr.ai/AqIc0
Follow-up on running #LLM locally: I benchmarked 4 models to see if I can actually work while they run
Previews toot: https://framapiaf.org/@lexoyo/116382060378966328
Good news: 3-7B models feel smooth, my laptop stays usable. The GPU handles most of the load.
The 20B model takes 4s before the first word — painful.
Sweet spot on my config: Lucie 7B, fast enough (19 tok/s) and good French.
Surprise: my system already swaps 2GB at idle — that's Firefox, not the AI 😅
Ivan Fioravanti ᯅ (@ivanfioravanti)
BenchLocal이 LM Studio와 4개의 병렬 실행을 이용해 테스트 중이며, 로컬 AI 생태계를 발전시키는 데 도움이 되는 도구라고 소개한다. 로컬 AI 벤치마킹/평가용 도구에 대한 주목할 만한 업데이트다.
Testei alguns modelos de #IA rodando localmente no #LMStudio para traduzir um PDF de 3 páginas.
Hardware:
i5-4460
RX 550 (4GB)
Configuração usada (igual para todos):
Contexto: 10130
GPU Offload: 24
Threads CPU: 2
Resultados
Gemma 3 1B → 25 segundos
Gemma 4 E4B → 3 minutos
GPT-OSS 20B → 11 minutos
Qwen 3.5 9B → 21 minutos
Conclusão
Modelos menores → muito mais rápidos
Modelos maiores → mais pesados e lentos
Dá pra rodar IA local mesmo em hardware mais antigo
If you are running #LocalLLMs you may be using LM Studio. Just a fair warning.... While this is practical, it's also proxying everything through their infrastructure. It's a privacy nightmare.
I tried out the Gemma AI models from Google, running locally on my AMD APU (Ryzen 7 Pro 7840U with Radeon 780M) and asked it some questions about ZFS send / receive.
gemma-4-26B-A4B-Q4_K_M:
14.29 tok/sec . The information, it generated, was factually correct and well laid out. Not the fasted, but surprisingly good.
gemma-4-E4B-Q4-K_M:
26 tok/sec. The information was completely wrong BS with made up parameters. The presentation was confident and well laid out. But it generated it quickly 😂
Bottom line: confidently incorrect at high speeds.
Habr
Moonstone2487
Stefan Malter | Malter365.de
LBHuston
Alex Hoyau
sayzard
Lord Daniel Salem
Leonid
Daniel S. Reichenbach
Larvitz 