.svg)
The Allen Institute for Artificial Intelligence (Ai2) is a non-profit research organization dedicated to advancing artificial intelligence through open, high-impact research. Operating under a truly open philosophy, Ai2 publishes models, datasets, and evaluation frameworks freely to empower the global research community and responsible AI development.
At the core of its work is the OLMo family of large language models, including Molmo for multimodal reasoning and Tulu 3 for instruction-following capabilities. Ai2 also provides an interactive playground at playground.allenai.org for hands-on access to its language and multimodal models, along with open datasets and evaluation frameworks designed to advance the science of AI.
Ai2 operates Asta, an agentic AI ecosystem purpose-built for scientific discovery. Asta combines autonomous research agents capable of exploring datasets and identifying research questions, a rigorous benchmarking suite called AstaBench, and a set of tools and standards for building and testing scientific AI systems.
On the environmental side, Ai2 deploys OlmoEarth, EarthRanger, and Skylight to transform planetary data into actionable insights for climate science, agriculture, wildfire management, and wildlife protection. Its embodied AI research develops intelligent robots trained in simulation environments to automate routine tasks.
Ai2 partners with the University of Washington, the National Science Foundation, and Google to amplify research impact. The institute actively recruits researchers and engineers through open roles and internship programs, fostering the next generation of AI scientists.
- Accessing and experimenting with open large language models via the Ai2 playground
- Using Molmo for multimodal reasoning tasks involving images and language
- Deploying Tulu 3 for instruction-following and alignment research applications
- Using Asta agents to autonomously explore scientific datasets and surface novel hypotheses
- Benchmarking AI research agents against rigorous standards using AstaBench
- Discovering scientific literature and research connections through Semantic Scholar
- Applying OlmoEarth models to climate science and environmental data analysis
- Using EarthRanger for real-time wildlife monitoring and conservation management
- Leveraging Skylight for maritime and environmental domain awareness applications
- Conducting foundational AI research using Ai2 open datasets and evaluation frameworks
- Developing and testing embodied AI systems using Ai2 simulation research tools
- Exploring open multimodal AI infrastructure for scientific discovery at national scale
