sayzardAISatoshi (@AiXsatoshi)
GLM-5-Q3와 Qwen3.5-397B-Q6 두 모델이 각각 약 320GB 수준이라며, 512GB 메모리 Mac이나 380GB 워크스테이션에서 적합한 크기인지 비교하려는 내용이다. 로컬/대형 모델 실운용 관점에서 메모리 요구량과 사용성을 비교하는 트윗이다.
Ajit Ananthadevanbeing a hardware guy, I've never written a realworld HTML page before, and I vibe coded this page :) https://simplycreate.online/services/
Although I know it's not perfect, was able to finish the job under 2 hours. After an initial niggle with some property name defined in my CMS theme. I was impressed it knew the theme inside out!
I'm happy it just flows with the rest of the theme, which was my main objective :)
Dragone2Migliori LLM locali del 2026: usali con Ollama o LM Studio
https://www.risposteinformatiche.it/migliori-modelli-llm-locali-2026-ollama-lm-studio/
N8 Programs (@N8Programs)
Qwen3.5-27B를 MLX 환경에서 4비트로 DWQ 처리한 결과를 공개했다. Qwen의 Int4 GPTQ 양자화를 기반으로 attention과 embedding 파라미터까지 4비트로 추가 양자화한 점이 핵심이며, 오픈소스 모델의 경량화·최적화 사례로 볼 수 있다.
Nicolas MØUARTQwen3.5-397B-A17B can't tell apart the differences between these 2 diamonds, leading to improper reasoning. ❌
Non Sequitur, it's safe to say that singularity has not happened yet. #Qwen #AI #LLM
Non Sequitur, it's safe to say that singularity has not happened yet. #Qwen #AI #LLM
HabrШаблон промта для улучшения взаимодействия с ИИ (на примере Qwen)
В статье предлагается практическое руководство для начинающих по работе с нейросетями. Вы узнаете как разрабатывать базовые промты, которые помогу твам избежать абстрактных и «водянистых» ответов от бесплатных ИИ-моделей. Представлены примеры качественных промптов, которые помогут улучшить взаимодействие с нейросетями и достичь более точных и полезных результатов.
https://habr.com/ru/articles/1013936/
#промптинжиниринг #промпт #эффективность #qwen #нейросети #обучение #составление_промпта #универсальные_промпты #нейросеть #шаблоны_для_работы
Özkan PakdilChecked out #Vulkan this morning, absolute beast. Then I tried installing OpenClaw one curl command and suddenly it wanted sudo root.
Now I’m reconsidering whether this setup is worth the trouble.
Anyway vulkan numbers here in case you want to run llama-server in an old laptop
https://ozkanpakdil.github.io/posts/my_collections/2026/2026-03-22-vulkan-llamacpp-debian-13/
Accelerating LLMs on Debian 13: Setting up Vulkan for llama.cpp
After setting up CUDA on my other laptop, I moved to a different(older) machine that doesn’t have an NVIDIA GPU. This one is an everyday laptop with integrated Intel graphics, but that doesn’t mean we have to settle for slow CPU-only performance. On this machine, I switched to the Vulkan backend for llama.cpp and the results were even more dramatic than I expected. Machine Hardware Info This laptop is running Debian 13 (Trixie/Sid) with the following specs:
金のニワトリ (@gosrum)
vibe-local에서 Nemotron-Cascade-2-30B-A3B(Q4_K_M)을 ts-bench로 평가한 결과를 공유했다. 결론은 Qwen3.5가 더 강하다는 내용으로, 로컬 추론 모델 성능 비교와 벤치마크 결과를 다룬 기술적 ट्वीट이다.
💫64기가💥👽몰루니움🖖
AlainDas chinesische #llm #MiniMax M2.7 übertrifft seinen Vorgänger M2.5 bei weitem und ist im Bereich von Athropics #Opus angekommen.
Leider scheint sich nun auch die Veröffentlichungsstrategie von #OpenWeight zu propritär zu wandeln.
Ich bin gespannt wie sich die chinesische Konkurrenz #Kimi, #GLM und #Qwen in Zukunft verhält.
Ich halte OpenWeight für unverzichtbar - der Markt vielleicht schon.
Leider scheint sich nun auch die Veröffentlichungsstrategie von #OpenWeight zu propritär zu wandeln.
Ich bin gespannt wie sich die chinesische Konkurrenz #Kimi, #GLM und #Qwen in Zukunft verhält.
Ich halte OpenWeight für unverzichtbar - der Markt vielleicht schon.