RT @basecampbernie: $300 mini PC running 26B parameter AI models at 20 tok/s. Minisforum UM790 Pro ($351) + AMD Radeon 780M iGPU + 48GB DDR5-5600 + 1TB NVMe. The secret: the 780M has no dedicated VRAM. It shares your DDR5 via unified memory. The BIOS says "4GB VRAM" but Vulkan sees the full pool. I'm allocating 21+ GB for model weights on a GPU with "4GB VRAM." The iGPU reads weights directly from system RAM at DDR5 bandwidth (~75 GB/s). MoE only activates 4B params per token = 2-4 GB of reads. That's why 20 tok/s works. What it runs: - Gemma 4 26B MoE: 19.5 tok/s, 110 tok/s prefill, 196K context - Gemma 4 E4B: 21.7 tok/s faster than some RTX setups - Qwen3.5-35B-A3B: 20.8 tok/s - Nemotron Cascade 2: 24.8 tok/s Dense 31B? 4 tok/s, reads all 18GB per token, bandwidth wall. MoE same quality? 20 tok/s. Full agentic workflows via @NousResearch Hermes agent with terminal, file ops, web, 40+ tools, all against local models. No API keys. Just a box on your desk. The RAM is the pain right now. DDR5 prices 3-4x what they were a year ago. But the compute is free forever after you buy it. @Hi_MINISFORUM @ggerganov llama.cpp + Vulkan + @UnslothAI GGUFs + @AMDRadeon RDNA 3. Fits in your hand. #LocalLLM #Gemma4 #llama_cpp #AMD #Radeon780M #MoE #LocalAI #AI #OpenSource #GGUF #HermesAgent #NousResearch #DDR5 #MiniPC #EdgeAI #UnifiedMemory #Vulkan #iGPU #RunItLocal #AIonDevice

Mehr auf Arint.info

#agent #API #GGUF #llama #LocalAI #OpenSource #Qwen3535 #arint_info

https://x.com/basecampbernie/status/2040326984446935059#m

https://winbuzzer.com/2026/04/04/google-previews-gemini-nano-4-android-aicore-developer-preview-xcxwbn/

Google Previews Gemini Nano 4 for Android AICore

#AI #Google #GoogleGemini #GeminiNano #Android #GenAI #AIModels #OnDeviceAI #EdgeAI #Smartphones #BigTech #Alphabet #OpenSourceAI

Google for Developers (@googledevs)

Gemma 4를 엣지 디바이스에서 활용해 챗봇을 넘어서는 AI 에이전트를 구축할 수 있다는 안내입니다. Gallery 앱에서 실험하거나 LiteRT-LM을 통해 노트북, 모바일, IoT 등 다양한 기기에 배포할 수 있어 온디바이스 AI 활용 사례를 넓힙니다.

https://x.com/googledevs/status/2040097383770689645

#gemma #edgeai #litetrlm #ondeviceai #iot

Google for Developers (@googledevs)

Google가 엣지 환경에서 실행되는 AI 에이전트 개발을 제안하며, Gemma 4를 #GoogleAIEdge로 활용해 노트북·모바일·IoT 등 다양한 기기에 배포할 수 있다고 소개했다. Gallery 앱에서 실험하거나 LiteRT-LM으로 배포할 수 있어 온디바이스 AI 개발에 유용하다.

https://x.com/googledevs/status/2039802553878094167

#google #gemma4 #edgeai #litetrlm #aiedge

Mekotronics R57-5S – Rockchip RK3576 mini PC and digital player integrates inclined 5-inch touchscreen display

https://fed.brid.gy/r/https://www.cnx-software.com/2026/04/02/mekotronics-r57-5s-rockchip-rk3576-mini-pc-and-digital-player-integrates-inclined-5-inch-touchscreen-display/

Toradex OSM and Lino SoMs – 30×30mm NXP i.MX 93/i.MX 91 modules with solder-down or B2B connector designs

https://fed.brid.gy/r/https://www.cnx-software.com/2026/03/31/toradex-osm-and-lino-soms-30x30mm-nxp-i-mx-93-i-mx-91-modules-with-solder-down-or-b2b-connector-designs/

I spoke with several ECE professors at #ECEDHA 2026 about how they’re integrating #AI into the curriculum. Everyone agrees it’s essential, but no one agrees on the best way to teach it. Check out my video interviews to learn more:

https://shawnhymel.com/3246/ai-in-engineering-education-lessons-from-ecedha-2026/?utm_source=mastodon&utm_medium=social&utm_campaign=ai_in_education

#engineer #education #edgeAI

Ultralytics (@ultralytics)

Ultralytics가 AxeleraAI와 파트너십을 맺고, Ultralytics Python 패키지에 네이티브 내보내기(export) 통합을 추가했다. 이를 통해 YOLO 모델을 Axelera의 Metis AI Processing Units(AIPUs)로 직접 배포해 엣지 추론용 생산 환경으로 빠르게 전환할 수 있다.

https://x.com/ultralytics/status/2038964748897726533

#ultralytics #yolo #axeleraai #edgeai #deployment

Rockchip RK3506J-based Forlinx FCU1501 fanless industrial IoT gateway offers dual Ethernet, plenty of I/Os and serial interfaces

https://fed.brid.gy/r/https://www.cnx-software.com/2026/03/31/rockchip-rk3506j-based-forlinx-fcu1501-fanless-industrial-iot-gateway-offers-dual-ethernet-plenty-of-i-os-and-serial-interfaces/