Intellect 1 is out. It's a 10B model trained across 3 continents using 100+ H100s, with 30 individual compute contributors. The evals are good (for 1T tokens), and the model is live. I can't stress how important this release is for open-source AI. Decentralized training is the only path toward sovereign open-source foundation models. This release proves that it's not just a fairy tale - it's working, and it's just the beginning.

RT @johannes_hage: great work! very interesting approach to initially limit the context length so the model learns to utilize it more effe…

@zack_overflow @Aryvyo I wish this program existed when I was a kid

@mike64_t 🤣🤣🤣

@mrsiipa @PrimeIntellect 😍

@taha_yssne Noicee

epic what ppl can do with a bit of compute

RT @LoubnaBenAllal1: We just published the second OpenR1 update with OpenR1-220k-Math, our new large-scale dataset for mathematical reasoni…

@AvpElk @GaryMarcus What does pure LLM even mean ? Again one should not make the confusion between architecture (LLM) and objective (next token prediction, rl ,...)

@Ar_Douillard It's a really well written paper

@TimDarcet @giffmana what if it was doing rl indefinitely ? As in the inference is just part of the RL process

@jonasgeiping amazing work

@aryaman2020 @jeremyphoward I was making the argument this approach would be the best way to scale reasoning. I have been proven wrong since tho

@aryaman2020 @jeremyphoward Does not mean that it's the end game 😉

@yaroslavvb @_arohan_ @KhonaMikail nice nice hehe its coming

@Dorialexander @dylan522p @ChrSzegedy But again, I am not picking a camp or trying to say who is right or wrong. But one needs to take the scaling part into perspective

@taha_yssne @Dorialexander this look sick, where can I try it

@Rixhbhco not yet !