Nathan

20M+ @liquidai LFM downloads, with a 1M downloads/week rate on Huggingface. Excited to see how our device foundation models, LFMs, empower a continuously growing ecosystem of on-device AI. 🐘

Today, we release LFM2.5-350M. Agentic loops at 350M parameters. A 350M model trained for reliable data extraction and tool use, where models at this scale typically struggle. <500MB when quantized, built for environments where compute, memory, and latency are constrained. 🧵

I started a part-time role at @ETH_AI_Center, mentoring students and working on post-training for the Swiss AI Initiative! 🤩Looking forward to working with interesting people like @a_yukh @ImanolSchlag @Noah_Xu_ @nathanrchn @ArnoutDevos If you are a student at ETHZ or EPFL looking for a semester or thesis project on post-training of LLMs, please reach out!

Yesterday I turned my MacBook Pro into a free AI employee. Not a chatbot. Not a coding assistant. An actual employee that writes my daily briefings, manages my house, and gets smarter every week. Here's the whole setup — no fluff, just what I actually did 🧵

I have a few qualms with the OpenAI API: For a Linux user, you can already build such a system yourself quite trivially by buying a 4xH100 box, installing it at home, installing CUDA and vLLM locally, and running GLM, Kimi, or a comparable open-source model. With typical consumer workloads, you should expect higher TPS for a fraction of the cost.