Our next generation of fully-open base and instruct models sit at the Pareto frontier of performance and training efficiency. Check out our [paper](https://arxiv.org/abs/2501.00656) to learn more, or keep reading for a summary.| allenai.org
