I'm happy it's released, it was such a cool post-training project Now I hope that app developers can take these models and build cool stuff with them, with less reliance on costly and slow cloud models. There's so much to do, life is too short to optimize ChatGPT prompts!
SOFTWARE
-
LFM2-1.2B Achieves Parity with Larger Models for Edge
By
–
For latency purposes on edge devices, we wanted it to be a non-thinking model. That was a big challenge, but we managed to squeeze a ton of performance from LFM2-1.2B and perform on par with much bigger models on our internal bench (see figure) but also BFCL v3 and v4.
-
350M Model Shows Power of Fine-Tuning for Big Data
By
–
It comes in two sizes: 1.2B and 350M I'm a big fan of the 350M model that can be used on GPUs to do big data operations. It's an absolute banger that shows how powerful fine-tuning can be
-
Tiny Task-Specific AI Models for Edge Devices
By
–
We're releasing a collection of tiny task-specific models Want to do data extraction, translation, RAG, tool use, or math on a Raspberry Pi? We got you covered! Here are a few examples ↓
-
OK Computer: AI Team Building Websites, Dashboards and Slides
By
–
pic.twitter.com/VyvR6wxG50 Kimi is cooking: OK Computer acts like a full AI product and engineering team, able to build websites, dashboards, and slides directly from chat. https://t.co/N4ANbYu1sD
— Chubby♨️ (@kimmonismus) 25 septembre 2025Kimi is cooking: OK Computer acts like a full AI product and engineering team, able to build websites, dashboards, and slides directly from chat.
-
India Empanels 14000 GPUs for Democratized AI Compute
By
–
AI has long been at the heart of India’s technology roadmap. In his earlier remarks, at the Rising Bharat Summit by @Network18Group, Shri @AshwiniVaishnaw spoke about empaneling 14,000 GPUs to drive democratized AI compute power.
— IndiaAI (@OfficialINDIAai) 25 septembre 2025
With India’s strong software base and… pic.twitter.com/wILEa4ztnOAI has long been at the heart of India’s technology roadmap. In his earlier remarks, at the Rising Bharat Summit by @Network18Group
, Shri @AshwiniVaishnaw spoke about empaneling 14,000 GPUs to drive democratized AI compute power. With India’s strong software base and -
Every Computer Science Subfield Now Oriented Toward AI
By
–
Every subfield X of computer science is now X for AI: hardware for AI, systems for AI, databases for AI, security for for AI, etc.
-
nvtop PCIe Generation Speed Monitoring Under Load
By
–
How does nvtop show the Pcie gen and speeds under load?
-
Stable Diffusion 3.5 FP8 TensorRT Checkpoints Now Available
By
–
Stable Diffusion 3.5 FP8 checkpoints, optimized for TensorRT, are now available. ⚡ 2x AI inference speed, half the VRAM, deployable on all NVIDIA Blackwell devices. Get started ➡️ nvda.ws/4gGkfii @StabilityAI
→ View original post on X — @stabilityai, 2025-09-24 23:01 UTC