A groundbreaking partnership between Lawrence Livermore National Laboratory (LLNL) and Meta has resulted in the creation of OPoly26, the world's largest open dataset dedicated to atomistic polymer chemistry. This massive collection of over six million quantum-accurate simulations is designed to accelerate the discovery of safer, faster, and more sustainable materials using artificial intelligence. Polymers are essential to modern life, forming the backbone of products ranging from clothing and packaging to advanced electronics and infrastructure. However, the development of new polymers has historically been hindered by a scarcity of high-quality data, particularly regarding complex behaviors like chemical reactivity. This gap is critical as the scientific community seeks to address environmental challenges, such as the recycling of plastics and the development of alternatives to Per- and Polyfluoroalkyl Substances (PFAS), often referred to as "forever chemicals." The OPoly26 dataset addresses these challenges by providing a diverse library of polymer structures with simulations performed at the density functional theory (DFT) level. Containing nearly ten times more data than the next largest comparable polymer dataset, it allows artificial intelligence models to learn patterns from millions of pre-computed structures in hours or days rather than years. This efficiency is pivotal for training machine-learned interatomic potentials (MLIPs), which predict how materials behave under real-world conditions. The collaboration leverages the unique strengths of both organizations. LLNL contributed significant computational power through its Tuolumne supercomputer and deep domain expertise in polymer science. This hardware enabled the team to compress years of simulation work into a manageable timeframe. In turn, Meta provided the vast computational resources necessary to execute 1.2 billion core hours of DFT simulations and trained state-of-the-art MLIP models, building on expertise gained from their earlier Open Molecules 2025 (OMol25) project. "This fills a huge gap," said Evan Antoniuk, a materials scientist at LLNL and co-principal investigator of the project. "We see this as a community resource, one that we hope becomes the go-to starting point for anyone interested in performing atomistic simulations of polymers." A key innovation of the OPoly26 dataset is its inclusion of reactive configurations. While many datasets focus on stable molecular structures, LLNL chemist Nick Liesen and LBNL co-principal investigator Sam Blau emphasize that capturing bond breakage and formation is essential for understanding polymer synthesis, manufacturing, aging, and recycling. By explicitly sampling hundreds of thousands of reactive scenarios, the dataset enables AI models to accurately describe the dynamic events that govern polymer behavior. Initial results indicate that incorporating this polymer-specific data alongside small-molecule training sets substantially improves model accuracy. The team has also introduced a suite of evaluation tasks to benchmark how well these models capture phenomena like polymer solvation. Future work will involve comparing AI predictions against experimental measurements to further validate the models. Rob Sherman, Vice President of Policy at Meta, highlighted the broader impact of the partnership: "By making this dataset publicly available, we're giving scientists potent new tools to address critical challenges in health care and beyond." With all data released under an open license, the researchers aim to maximize reuse and reproducibility across academia, industry, and government. This open science approach ensures that the public-private investment yields broad benefits, democratizing access to the high-fidelity data needed to revolutionize materials design.