sayzardArtificial Analysis (@ArtificialAnlys)
Artificial Analysis에서 Gemma 4, Qwen3.5 등 여러 AI 모델을 비교할 수 있는 모델 비교 페이지를 소개했다. 최신 오픈 모델들의 성능을 한곳에서 확인하고 벤치마크 비교에 활용할 수 있는 유용한 리소스다.
ThomasNew week, new update for the slides of my talk "Run LLMs Locally":
Now including Gemma4 and Qwen3-Omni with Vision and Audio support and new slides describing Llama.cpp server parameters.
https://codeberg.org/thbley/talks/raw/branch/main/Run_LLMs_Locally_2026_ThomasBley.pdf
#ai #llm #llamacpp #stablediffusion #gptoss #qwen3 #glm #localai #gemma4
Techgedöns 👨🏼💻📱 #TchgdnsVoicebox: #OpenSource-Alternative zu #ElevenLabs auf Basis von #Qwen3-TTS • Voice Cloning und Text-to-Speech #KI 🔊🤖✨ https://tchgdns.de/?p=162038
Hublai (charismatic megafauna)levelup.gitconnected.com/i-tested-the...
#PrismML 8.2 billion parameters in 1.15 GB competes with #Llama3.1, #Qwen3, and #Gemma4 FP16 models in16 GB. PrismML’s Bonsai 8B is 14x smaller. On iPhone 17 Pro Max, it clocks 44 tokens per second: real-time conversation speed on a phone, no cloud required.
I Tested the 1-Bit LLM That Fi...
I Tested the 1-Bit LLM That Fi...
Omar Khattab (@lateinteraction)
@a1zhang의 새 블로그가 언어 모델의 미래를 다루며, RLM-Qwen3-4B에 대해 32k 토큰의 쉬운 장문맥 과제로 GRPO를 학습해도 1M 토큰, 8-needle 장문맥 작업으로 자동 일반화되고 100% 신뢰도로 동작한다는 결과가 핵심으로 소개됐다.
Omar Khattab (@lateinteraction) on X
New must-read blog by @a1zhang on the future of language models. Buried nugget: doing GRPO for RLM-Qwen3-4B on short (32k token) and easy (single-needle) MRCRv2 long-context tasks generalizes *automatically* and with perfect (100%) reliability to 1M-token, 8-needle tasks!!
William Ruider (@ruider92545)
EXO Labs 1.0.69이 2022년형 6노드 Mac M1 Studio Max 클러스터에서 Thunderbolt 4와 MLX ring만으로 대형 모델 Qwen3.5-122B-A10B(8-bit/FP8, 131GB)를 구동하는 놀라운 성능을 보여줬다는 내용이다. 로컬 분산 추론/실행 성능의 큰 진전을 암시한다.
William Ruider (@ruider92545) on X
Guys and Girls, look >>> what EXO Labs 1.0.69 did to my 6-node Mac M1 Studio Max cluster (from 2022) over Thunderbolt 4 without RDMA, MLX ring. Just downloaded Qwen3.5-122B-A10B (8-bit/FP8) MLX 131 GB heavy. Thinking mode I have a …. WHAT!!!! No way!!!! Am I dreaming ???!
Axel RafnNew update for the slides of my talk "Run LLMs Locally": Bonsai-8B
The latest version of Llama.cpp now supports Vulkan with 1-bit quantized models like Bonsai: 8B model having 1.1 GB in size, 2.5 GB in RAM.
https://codeberg.org/thbley/talks/raw/branch/main/Run_LLMs_Locally_2026_ThomasBley.pdf
#ai #llm #llamacpp #stablediffusion #gptoss #qwen3 #glm #localai
Chi Kim🤯 It's a dumb experiment but cool at the same time! I asked an #OpenClaw #agent inside a Docker container with browser access (powered by #Qwen3.5-27B on a local machine) to create its own email account. It tried Proton, Yahoo, GMX, Mail.com, Zoho, etc..., but kept failing bc phone verification. It kept going at it for about three hours until it successfully created an account with Tutanota by itself. I was able to verify the account using the login info it gave me. #LLM #AI
金のニワトリ (@gosrum)
Gemma-4를 thinking 없이 실행해 본 결과 성능이 떨어졌고, Qwen3.5와 달리 ts-bench 점수가 낮아졌다고 합니다. 성능을 중시한다면 Gemma-4는 기본값인 thinking 모드로 사용하는 것이 좋다는 실전 평가입니다.