sukjong Yi retweeted
これはCAD界隈ではかなりの大ニュース、、!!
NVIDIAが持っているCADのエンジンSMLIBが、本日オープンソースで公開されました。しかも商用利用OK。
引用元は英語なので補足すると、20年以上CADを作ってきた海外の会社の方が、CADにとってビッグニュースだと喜んでいる投稿です。
・CADの心臓部(曲面や立体を計算するエンジン)がタダで使える
・これまでは既存の契約企業しか使えなかった
・AIから直接動かせるPython版も無料で公開済み
AIは図面を読むのは得意でも、描くのはまだ少し弱かった。そこにプロ用のCADエンジンを無料でつなげられるのはデカい。
手描きスケッチからCADデータを起こすような下書きが、一気に現実的になりそう!
Big news for CAD: SMLIB, NVIDIA's NURBS based kernel is finally open source today 🍻🎉
We have been working with SMLIB and its dev team for more than 20 years. It enables all our products: classcad.ch headless CAD, GUI, AI/MCP, PDM, etc
github.com/NVIDIA-Omniverse/…
sukjong Yi retweeted
From 3.7B to 36B, whatever size a developer needs, the best model in that band is a K2-Horizon: the smartest at its size, the smallest at its level of intelligence, and interchangeable with the rest of the fleet.
K2-Horizon-MoVA-36B-A4B, K2-Horizon-7B, and K2-Horizon-3.7B all sit on the Pareto line, delivering the best cost per intelligence across model sizes. Built as one fleet, they share the same vocabulary and chat templates, so teams can choose the right model for each workload without changing how they build.
Specially:
• K2-Horizon-MoVA-36B-A4B: Matches dense 27B models on 4B active parameters per token.
• K2-Horizon-7B: Beats dense open models three times its size.
• K2-Horizon-3.7B: Best model under 4B
All available on Hugging Face under Apache 2.0:
• K2-Horizon-MoVA-36B-A4B: huggingface.co/IFM/K2-Horizo…
• K2-Horizon-7B: huggingface.co/IFM/K2-Horizo…
• K2-Horizon-3.7B: huggingface.co/IFM/K2-Horizo…
sukjong Yi retweeted
🔥 Tencent released an 8.5GB model that can VISUALLY 👀 search your private PDFs, charts, tables, and scanned documents, all done completely locally.
EVIE-Preview-4.5B is not a chatbot model.
🎯 It's a visual document retrieval model built on Qwen3.5-4B that searches the actual page, including layout, charts, tables, images, and text, instead of flattening everything into text chunks first.
And this tiny 4.5B model is taking on much larger models. 👀
🏆 #1 on ViDoRe V3
🏆 #1 on ViDoRe V1+V2
ViDoRe V3:
🥇 EVIE 4.5B → 65.36
✧ webAI ColVec 8.4B → 65.32
✧ NVIDIA Nemotron 8B → 63.54
✧ Jina Embeddings v4 → 57.54
Stats ...
🧠 4.54B parameters
📦 ~8.5GB BF16
🔎 128-dimensional visual token embeddings
🌍 Multilingual
📜 Apache 2.0
💻 Runs locally
You can give your local AI thousands of these ...
📑 PDFs
📊 Charts
🧾 Scanned forms
📚 Research papers
📈 Financial reports
🛠️ Technical manuals
…and it can retrieve the actual pages visually, not just OCR text. 🎯
This feels like an important next step for private Local RAG.
⚠️ No llama.cpp / Ollama / LM Studio support yet. The current path is PyTorch + ColPali.
🔗 HF /tencent/EVIE-Preview-4.5B
sukjong Yi retweeted
The harness can make a difference in AI for CAD.
Below: same LLM (Opus 4.8) with different harnesses.
Looking forward to see more progress without waiting for a new LLM version :)
(Good job @pzfreo)
sukjong Yi retweeted
卧槽,终于有人对CAD行业下手了
过去几年,我们见到了很多3D的AI模型,但是那些都是玩具,很难用,基本无法用到到工业生产
但刚刚麻省理工开源了一个将照片转换为可完全编辑的 CAD 程序的 AI 模型GenCAD,它悄然终结了每小时 150 美元的 CAD 建模行业。
GenCAD不只是生成一个3D外形。
它做的是:给一张CAD图像/渲染图,生成对应的参数化CAD命令序列,也就是CAD program,再通过几何内核转成3D solid model。
重要的是它保留了「建模历史」。
所以它不是「AI画了个3D玩具」,而是在尝试把图片变成可编辑、可制造流程能继续接住的CAD模型。
同时它:
→ 无需 SolidWorks 许可
→ 无需数周建模
→ 不需要 CAD 工程师
100% 开源
如果后续这条路线跑通,CAD行业最贵的那部分时间,草图理解、参数化建模、反复改模型,都会被一点点吃掉。
代码、数据和预训练模型入口都已经放出来了。
GenCAD可能是AI真正进入工业设计工作流的一个信号。
项目地址:gencad.github.io/
视频Demo:gencad.github.io/static/vide…
Paper:arxiv.org/abs/2409.16294
github:github.com/ferdous-alam/GenC…
MIT open-sourced an AI model that converts photos into fully editable CAD programs and it quietly kills the $150/hour CAD modeling industry.
Just upload a sketch or photo and it generates the full parametric 3D model. exportable as STL. ready for manufacturing.
→ no SolidWorks license
→ no weeks of modeling
→ no CAD engineer needed
100% Open Source
sukjong Yi retweeted
Bu çok iyi 🚨 bir şehri 3D haritaya çevirmek artık çok daha kolay.
**map3d** adlı açık kaynak proje, OpenStreetMap verilerini kullanarak gerçek dünyadaki binaları, yolları ve yapıları fotogerçekçi 3D modellere dönüştürüyor. Üstelik GLB formatında dışa aktarma desteği de var.
Şehir görselleştirme, dijital ikiz projeleri, 3D prototipler ve kentsel planlama için oldukça güçlü ve kullanışlı bir araç.
Demo’yu görünce etkilenmemek elde değil.
🔗 Demo: map.fleet.im/
🔗 GitHub: github.com/cartesiancs/map3d
Sizce map3d gibi açık kaynak 3D harita araçları, şehir planlaması ve dijital ikiz projelerini ne ölçüde hızlandıracak?
Düşüncelerinizi yorumlara yazın, konuşalım. 🗺️
sukjong Yi retweeted
KAIST, 리튬금속 배터리 '망가지는 순간' 포착…성능↓ 원인규명
yna.co.kr/view/AKR2026050806…
OpenCode x Ring 2.6 1T - free for a limited time
256K context • reasoning • text only
Thanks to @AntLingAGI and @novita_labs for making the model available
EMO is a 1B-active, 14B-total MoE trained on 1T tokens with 8 of 128 experts active per token.
It matches standard MoEs on general benchmarks but allows selective expert use: for each domain, we deploy only the most relevant experts while keeping full-model performance.
sukjong Yi retweeted
We are launching Ring-2.6-1T, a trillion-parameter flagship thinking model engineered for real-world complex tasks and production env: 🚀
- Adjustable Thinking Effort: dynamic compute mechanism to flexibly balance cognitive depth, token cost, and execution speed;
- Agent-Optimized: Built for high-frequency workflows, delivering rapid multi-step execution and tool orchestration with SOTA stability;
- Deep Thinking: Unlocks the model's maximum capability ceiling for rigorous mathematical logic and scientific research;
sukjong Yi retweeted
UPDATE MAY 7: Scaling work continues on a slope above SR 20 near Diablo Lake. Crews use hand tools, air compression (and a little muscle) to send loose rock down the slope.
SR 20 remains closed from MP 130-156. More info in our latest blog update bit.ly/SR20blog
sukjong Yi retweeted
지난 4일 경북 울진군 북면 하당리에 건설 중인 송전탑(21호기) 아래에서 송재순 산림기술사가 무릎을 꿇고 경사면 바닥에 바짝 엎드렸다. 10m가 넘게 이어진 틈 사이로 손을 집어넣자 팔꿈치까지 쉽게 잠겼다.
khan.co.kr/article/202605080…
Codex now works directly in Chrome on macOS and Windows.
It’s even better at working with apps and sites in Chrome, and now works in parallel across tabs in the background without taking over your browser.
To get started, install the Chrome plugin in the Codex app.
We are releasing ZAYA1-8B open-weights under Apache 2.0, free to try on Zyphra Cloud today.
Blog: zyphra.com/post/zaya1-8b
Technical report: zyphra.com/zaya1-8b-technica…
Weights: huggingface.co/Zyphra/ZAYA1-…
Available on Zyphra Cloud: cloud.zyphra.com
sukjong Yi retweeted
兄弟们!Google 刚刚更新了Code Wiki!
Google 直接把开发者最头疼的“读代码”瓶颈,一键干掉了。
他们刚刚发布的 Code Wiki,让任何一个代码仓库Github都能瞬间拥有一个永远最新、结构化、 智能可聊的 Wiki。
核心三招狠活:
- 自动扫描仓库,代码一改文档就实时更新,再也不用担心文档过时
- Gemini 驱动的智能聊天,直接把整个 Wiki 当知识库,回答问题还能超链接跳到具体代码行
- 自动生成架构图、类图、时序图,点一下就能从解释直达源码
新手第一天就能上手,老代码没人维护也能秒懂,大库小库通吃。
Google 自己都说:“读现有代码是软件开发里最大、最贵的瓶颈之一。”
现在,这个时代终于要结束了。
公共仓库已开放预览:codewiki.google
你觉得 Code Wiki 会不会成为下一代代码理解标配,把“读代码”从痛苦变成享受?
sukjong Yi retweeted
Autodeskが無料公開!数千種類のアセットを置くだけで3Dモデルが作れる「Project Falcon」
modelinghappy.com/archives/6…
✔ 数千のパーツを配置して3Dモデル作成可能
✔ スナップ・グループ・ミラーなど直感操作に対応
✔ ブラウザ保存でいつでも再開OK
✔ 書き出してBlenderやMayaでも使用可能
✔ 無料&初心者でもレゴ感覚で作れる
이제 캐드 LLM모델도 나온다.
- 핵심 프로젝트 (Zero-To-CAD): 최근 가장 활발한 프로젝트로, 이미지를 텍스트(CAD 프로그램 코드)로 변환해주는 Qwen3-VL-2B 모델과 이를 위한 대규모 데이터셋(100k, 1m 규모)이 포함되어 있음.
- 모델 (17개): 이미지에서 3D 형상을 추출하는 'Make-A-Shape', 텍스트를 이미지로 바꾸는 'WaLa' 시리즈 등이 있음.
- 오토데스크랩에서 공개함.
huggingface.co/ADSKAILab
sukjong Yi retweeted
Introducing ERNIE 5.1 Preview — now live! 🚀
Ranked #13 globally and #1 among Chinese labs on @arena 's Text Arena.
Top-10 worldwide across:
📐 Math — #9
⚖️ Legal & Government — #1
💼 Business, Management & Financial Ops — #4
💻 Software & IT Services — #7
Built on the strong pre-training foundation of ERNIE 5.0, ERNIE 5.1 Preview compresses total parameters to ~1/3 and activated parameters to ~1/2 of its predecessor, while using only ~6% of the pre-training cost of comparable models — delivering leading foundational performance at its scale.
Try it now 👉 ernie.baidu.com/
More new models are on the way — stay tuned 👀









































