Research
Our mission is to build an open platform to fuel and accelerate AI breakthroughs globally. Our models, robust tooling, reproducible pipelines, and evaluation frameworks are designed for collaboration and contribution, empowering others to build, iterate, and innovate faster.
AllResearch PaperDatasetBlog
| Title | Category | Date Published |
|---|---|---|
| Dataset Reaches #1 on Hugging Face | Dataset | June 2025 |
| Muon Doesn’t Clearly Grok Faster | Blog | June 2025 |
| ESSENTIAL–WEB V10: 24T tokens of organized web data | Research Paper | June 2025 |
| Practical Efficiency of Muon for Pretraining | Research Paper | May 2025 |
| Muon, muP, and the Compute–Time Tradeoff | Blog | May 2025 |
| Rethinking Reflection in Pre-Training | Research Paper | April 2025 |
| Rethinking Reflection in Pre-Training | Blog | April 2025 |
| Layer Sharding for Large–Scale Training with Muon | Blog | May 2025 |