More AI breakthroughs at @essential_ai from the incredible team @ashVaswani has been building. More to come...
Data is the new fabric. The richer the dataset, the more vibrant the creation. This unlocks new dimensions in ME+AI.
Brilliant! @ashVaswani goes legend again by releasing a massive, fully labeled dataset that makes training easier.
Excited to see people train SOTA models with Essential-Web v1 and @AMD Instinct GPUs.
Explore the dataset here- huggingface.co/datasets/Esse……
👏 Congrats to @andrewhok, Michael Pust, @timr1126 & the brilliant Data team at @essential_ai for your commitment to open data curation.
Thank you for including @AIatAMD in your open science journey.
Check out the most meticulous open-sourced foundation model data. Amazing work by the team!
@ashVaswani and his team @EssentialAI are cooking 👨🍳
data team @essential_ai has been cooking! we just shipped a 24T–token corpus of richly-labeled CC-based data, significantly lowering the bar to entry for curating web-scale targeted datasets
s/o @AndrewHojel Michael Pust @RitvikKapila @ashVaswani
datasets unlock the future!
I am reminded of this phenomenal read from @RenPhil21 renaissancephilanthropy.org/…
Check out our recent work at @essential_ai; Essential-Web v1.0- a web scale corpus of 24T tokens which we find useful to curate high-performing domain specific datasets for LLM pre-training.
Paper link: arxiv.org/abs/2506.14111
cc @AndrewHojel @timr1126 @YashVanjani @ashVaswani
@ashVaswani doing gods work over here and making the basics much better. Bigger and better quality means better models. I wonder what he should use to embed to deal with all those tokens.
Very strong result @ashVaswani !