Qwen3.5-4B-3D-Coder w/ Three.js A 4B parameter model can generate interactive 3D scenes using Three.js. Qwen3.5-4B-3D-Coder's October 29 update makes the full FP16 weights, training dataset, and technical report publicly available. The dataset contains 4,146 synthetic examples. Training took about 60 hours, combining full fine tuning with a short LoRA stage. 3D scenes are generated as HTML and JavaScript through Chat mode, not agent tool integration. Specific generation settings and a Three.js system prompt are required. Q4_K_M and Q8_0 GGUF versions are also available, with more quantized variants planned. huggingface.co/ccore/Qwen3.5…

Maya v1.1.42 is out Vision enabled on every model and every build 👁️‍🗨️👁️‍🗨️ DeepSeek reads long first prompts 16-24% faster, thanks to @needmorevram. github.com/mw00/project-maya…

vllm-radiance, a token monster!!! running vllm-radiance with 2x AMD R9700. try it if you just got same rig. codeberg.org/StillDeadcode/v…

The iteration of local inference is move so fast. The benchmark freshed day by day. But at least, if you want to play, you should get a card with 16GB VRAM. Yea, cuz we have no smaller model with performance of qwen3.8 27b. right?

FPGA data base updated with CoreScore, and the design tools available including open source. thegatearray.com

Local inference just opening a new world. There is one day, you try to install an app, it says "sorry, you need at least a 16GB VRAM card!". Just like you try to install a game. 😀

Intel B-series cards are now fully supported 🥳 + Overall improvements & fixes + Some tweaking on R9700 Next up is Strix, Mac, parallelism and PERFORMANCE 👀 See more at github.com/Niko1221/Strata/r…

Physics is hard to visualize through static images alone, so I turned my 174 pages of high school physics notes into an interactive website. It covers mechanics, electricity, and magnetism, breaking down complex concepts into intuitive visualizations, real-world scenarios, and interactive problems. Check it out here: atlas-of-physics.vercel.app/

Introducing, 100% free charts kit for you guys. Spent whole damn month working on this 62 chart components. Available in Figma, code, or connect your MCP. Download here: mebius.design/freebies/chart…

.@TinyGolang announces TinyGo 0.44 with smaller and faster programs, more multicores on microcontrollers, support for Arduino UNO R4, new drivers, and big Bluetooth updates! tinygo.org/blog/2026/tinygo-…

Well, I have to say, the fan create more noise compare to Intel ARC B70. even they are all turbo style, but the b70 is so quiet just like other game cards. For token preformance from my experience: Radiance engine is the best for amd r9700 by now. XeStrata engine is the best for intel arc series by now. Strata, do well for resource limited rigs, but for it's just a bit short of performance with certain rigs. Just like all of those.

🚀 Introducing r9k — our from-scratch LLM inference engine at Unassuming AI. Built from the ground up in C++ and HIP, optimized for Qwen3.8-27B on just 2× AMD Radeon AI PRO R9700 GPUs. ⚡ 127.9 tokens/s single-stream decoding 🔥 2.6× faster than our tuned vLLM baseline 🚀 264.6 tokens/s with 8 concurrent requests ⏱️ 480ms time to first token Under the hood: • Custom GPU kernels • FP8 inference • Lossless speculative decoding • GPU graph execution • Custom tensor parallelism No CUDA. No NVIDIA GPUs. Just AMD, HIP, and a lot of low-level optimization. Same model. Same hardware. Same benchmark harness. We’re building faster, more efficient inference infrastructure — from the ground up. 📄 Technical report: unassuming.ai/report/r9k-tec… 🌐 Try it out: unassuming.ai #LLM #Inference #AMD #ROCm #AIInfra

r9k w/ 2 x AMD Radeon AI PRO R9700 Qwen3.8-27B Unassuming AI's r9k reaches 127.9 tps single stream decoding Qwen3.8-27B on 2 x AMD Radeon AI PRO R9700 GPUs. Built from scratch in C++ and HIP, the engine is reportedly 2.6x faster than the team's tuned vLLM baseline on the same hardware and benchmark harness. With 8 concurrent requests, aggregate throughput reaches 264.6 tps. Time to first token is 480ms. It uses custom GPU kernels, FP8 inference, lossless speculative decoding, GPU graphs and custom tensor parallelism without CUDA.

DGX Spark support is now official! DeepseekV4.1-Flash-1017 on Mac! Updates are incoming for Intel B-series cards & Strix🫡

📚 Highly Cited Paper in JLPEA! 📄 “A Low-Power 5-Bit Two-Step Flash Analog-to-Digital Converter with Double-Tail Dynamic Comparator in 90 nm Digital CMOS” 👥 Authors: Reena George and Nagesh Ch 🔗 mdpi.com/2079-9268/14/4/53 #LowPowerDesign #AnalogToDigitalConverter #CMOS #IoT

20 Open-Source Humanoid Robots You Can Build Yourself | From $50 to $900 What if you could build your own humanoid robot without spending thousands of dollars? In this video, I explore 20 open-source and DIY humanoid robotics projects, ranked from 1 to 20. The list covers a wide range of budgets, from around $50 to $900 for an upper-body build. Some projects provide official 3D models, CAD files or assembled STEP models. Others do not have a complete, ready-to-use 3D model matching the robot shown in the video. For some entries, I arranged available components or created my own visual representations. The 3D models for projects 2, 8, 10 and 6 require verification. Some visuals may be reconstructions or assemblies rather than official models of the original robots. Please check the project documentation before treating any visual as an exact representation of the hardware. The 20 projects featured 1. RoboPrime | Around $60–100 A low-cost robotics project with an open-source design. 2. Open-source DIY humanoid builds A general category covering custom humanoid builds assembled from different hardware components. The humanoid shown is a generic 3D model created by me, not an official robot model. 3. Qmini | Price unconfirmed A compact robot project from Unitree Robotics. 4. Zowi | Around €100 A small educational robot associated with BQ. 5. Otto Humanoid / Otto DIY | From around $50 A DIY robotics platform. The final cost depends on the selected components and configuration. 6. Kayra | Price unconfirmed The visual combines official leg designs with additional components positioned by me. 7. PLEN2 | Around $499 for the kit A compact humanoid robot with published CAD files and walking-related software. 8. Nova Desktop Humanoid | Around $150 A desktop humanoid project by Daniel Hany. I positioned the available head and arm components myself in the 3D scene. 9. Bambot | Around $300 A low-cost robotics project with open-source design resources. 10. Rapiro | Around $390 A small humanoid robot designed by Shota Ishiwatari. The 3D reconstruction shown in the video requires verification against the original model. 11. Valera | Around $400 An open-source humanoid robotics project. 12. Reachy Mini Lite | Around $399 A compact robotics platform developed by Pollen Robotics in collaboration with Hugging Face. 13. Open Duck Mini | Under $400 An open-source bipedal robot project for robotics enthusiasts and developers. 14. MicroDuck | Around $399 A Pollen Robotics project with reinforcement-learning resources. 15. ROBOTIS MINI | Around $540 A compact humanoid platform associated with ROBOTIS components. 16. Halley | Around $500–800 An open-source robot project with 3D design resources. 17. Microban | Around $550–600 A robotics project associated with the Rhoban team at the University of Bordeaux. 18. Endeavor II | Around $600 A humanoid project with an assembled STEP model. 19. XLeRobot | Around $660 An open-source robotics project with accessible hardware designs. 20. InMoov | Around $900 for the upper body Created by Gaël Langevin, InMoov is a well-known 3D-printed humanoid robotics project. What I noticed while preparing this video The projects vary in price and in the amount of design information available. Some provide official 3D models and detailed mechanical files, while others need additional work to create a visual representation. The models shown for projects 2, 6, 8 and 10 need to be distinguished from verified official 3D assets. Check the original repositories and documentation for confirmation.

【エルミタ的速攻撮って出しレビュー Vol.1703】12V-2×6の給電トラブルを監視する、Noctuaファン搭載のThermal Grizzly「WireView Pro II」とは gdm.or.jp/review/2026/1026/6…

コードを1行も書かず、64ビットRISC-VでLinuxが動いた。 Claude Codeが38日間で設計した「mmRISC-2」。RTL 22,100行、8段パイプライン、FPGAでDebian起動。@GlobalFoundries 180nmも視野に。 圓山宗智さんが11/11 東大本郷で語ります。 詳細 riscv.or.jp/mmrisc2/ 参加登録 peatix.com/event/5190001/ #RISCVDayTokyo @risc_v

ちょっと古い投稿ですが紹介 vLLM Radiance、R9700x2でQwenFNのDecodeが176t/s prefillが6450t/s Strataのスコアと同じか、早いぐらいです。 Strataの音速バージョンが公開されるまではこちらに住んでいようかな？ reddit.com/r/LocalLLaMA/comm…

ICY DOCK、2台のE1.S SSDをリムーバブルドライブ化するPCIe 5.0拡張カード ➡️サーマルスロットリングを抑える50mmの冷却ファンを搭載 gdm.or.jp/pressrelease/2026/…