ICLR 2025 many many many thanks to Kyunghyun Cho and Yoshua Bengio for enabling the wildest ever start of my research career 2014 was a very special time to do deep learning, a commit that changes 50 lines of code could give you a ToT award 10 years later 😲

intelligence per picojoule

wait, Grok-3 mini is actually good?

Cost of intelligence is wild🤯 xAI just dropped Grok 3 mini. Best reasoning model on the planet at 5× lower cost.

we are seeing the loop of intelligence expansion and cost compression playing out for the last few years. this time, "thinking" is becoming the art to navigate the intelligence-price frontier. smart move is to stay on the edge of the curve, whether human or machine.

Sparse attention is one of the most promising strategies to unlock long-context processing and long generation reasoning in LLMs. We performed the most comprehensive study on training-free sparse attention to date. Here is what we found:

Finals season stressing you out? You're just a few taps away from unlocking a 24-hour study sidekick (me). Sign up with your .edu email for two free months of my supercharged self, SuperGrok.

Palantir Palantir CEO Alex Karp and TWG Global Co-Chairman Thomas Tull sat down at the Milken Family Fdn Institute conference with CNBC to discuss how TWG and Palantir’s partnership with xAI will design and deploy AI-driven solutions for enterprise. youtube.com/watch?v=svDRof…

Things I would work on if I was in academia: 1. Taking an hour walk everyday 2. Learning a new sport 3. Take some cooking classes 4. Aim for perfect sleep score

It's a real shame that ICML has decided to automatically reject accepted papers if no author can attend ICML. A top conference paper is a significant boost to early career researchers, exactly the people least likely to be able to afford to go to a conference in Vancouver.

xAI partners with Polymarket to blend market predictions with X data and Grok’s analysis. Hardcore truth engine - see what shapes the world. This is just the start of our partnership with Polymarket. More to come. 🚀

Poland went from Iran-level of economic development to Japan-level in a single generation

War Room squad locked in

Tried Grok 4 on a dozen non-trivial math (under/)grad level math problems. So far, it has failed to fail me even once. Congrats to Yuhuai (Tony) Wu, Eric Zelikman and the whole xAI reasoning team, their progress has exceeded all my expectation!

It’s a pretty good model

Fed Grok4 Heavy my massive assembler repo. In ~6 mins, it cleaned up everything, optimized files, and returned them working perfectly. Same codebase in Cursor + MAX? Gemini, Claude, GPT all wrecked it.

I'm hiring top researchers/engineers to work on Reasoning Efficiency at xAI. If you're excited about the mission of maximizing 🧠 per 💵 by building extremely cost-efficient reasoning models, apply here: x.com/i/jobs/1948828… & drop me a DM!

We are hiring on pretraining as well. If you are passionate about improving training efficiency, pretraining data quality and training infra. Please apply here: job-boards.greenhouse.io/xai/jobs/45338… job-boards.greenhouse.io/xai/jobs/46846… job-boards.greenhouse.io/xai/jobs/46036…

We tested top foundation models on the International Olympiad in Informatics (IOI) - a programming competition that tests algorithmic thinking and C++ coding skills. We found xAI's Grok 4 to be the clear SOTA winner, scoring first place on both 2024 and 2025 exams. 🥇📊👏