More ways to scale TorchTitan with SkyPilot great to see the community expanding the ecosystem with new tutorials like this one!
OPEN SOURCE
-
Open Weights Models Secured Against Permanent Loss
By
–
open weights already backed up, boy will never retire
-
LFM2-8B-A1B GGUF Quantizations Available on Hugging Face
By
–
LFM2-8B-A1B and GGUF quants are now available on Hugging Face!
-
LFM2-8B-A1B: Efficient MoE Language Model Released
By
–
LFM2-8B-A1B just dropped on @huggingface!
— Maxime Labonne @ ICLR (@maximelabonne) 7 octobre 2025
8.3B params with only 1.5B active/token 🚀
> Quality ≈ 3–4B dense, yet faster than Qwen3-1.7B
> MoE designed to run on phones/laptops (llama.cpp / vLLM)
> Pre-trained on 12T tokens → strong math/code/IF pic.twitter.com/cGbuJoOMDNLFM2-8B-A1B just dropped on @huggingface
! 8.3B params with only 1.5B active/token > Quality ≈ 3–4B dense, yet faster than Qwen3-1.7B
> MoE designed to run on phones/laptops (llama.cpp / vLLM)
> Pre-trained on 12T tokens → strong math/code/IF -
DPT-2 Document Pre-Trained Transformer Model Revolutionizes Complex Table Parsing
By
–
The new Document Pre-Trained Transformer (DPT-2) model in ADE is a game changer. It parses complex tables with merged cells, multi-level headers, and irregular grids, with spatial grounding and bounding boxes for traceability. GitHub Repo:
-
ade-python: Transform Complex Documents into LLM-Ready Data
By
–
Turn complex and messy documents into LLM-ready data! ade-python is a Python library for Agentic Document Extraction (ADE) that outputs layout-aware structured JSON from visually complex documents.
-
Can startups survive building on LLMs against big tech?
By
–
So based on the trend, can we say that building tools on top of LLMs is a bad idea? Eventually you will get killed by big companies? Probably only open sourced alternatives will survive.
-
University of Zurich Joins Hugging Face Academia Hub
By
–
Super happy to announce that University of Zurich (
@UZH_en
) just joined @huggingface Academia Hub Their students and educators get a better access to collaboration and compute features (including ZeroGPU power) on the Hub. -
Open vs Closed Models: Gap Narrowing, Focus Shifts to Applications
By
–
The gap between open and closed models are narrowing and I expect this trend to continue. As foundation models become commoditized on a global level, the most interesting directions from both research and commercial is not in their development but in finding new ways to use them!
-
Open Source AI: Democratization Over Restrictions
By
–
In 2024 I wrote this opinion piece for Level: My opinion about open source AI Summary: Open-source AI is risky, but banning it is worse. It democratizes knowledge, removes single points of failure, and accelerates civilization’s progress. Proprietary AI enables