Ai2 participated in TMax, an open research project developing reinforcement learning recipes for terminal agents. The paper identifies work performed while Hamish Ivison and Nathan Lambert were at Ai2 and credits the institute’s Beaker team with infrastructure help.
The authors acknowledge Nebius cloud credits that helped run evaluations for the project. This disclosure establishes research evaluation use. The paper also describes model training and sandbox infrastructure, but does not attribute that training infrastructure to Nebius.
TMax: A simple recipe for terminal agents — second version
Listed 19 Nov 2026 · arXiv · Web page
The TMax authors credit Ai2’s Beaker team with infrastructure help and Nebius cloud credits with supporting research evaluations. The second version’s arXiv publication header is dated 2 November 2026.
TMax: A simple recipe for terminal agents — second versionListed 19 Nov 2026 · arXiv · Web pageThe TMax authors credit Ai2’s Beaker team with infrastructure help and Nebius cloud credits with supporting research evaluations. The second version’s arXiv publication header is dated 2 November 2026.
Ai2 (the Allen Institute for Artificial Intelligence) is a Seattle-based nonprofit research institute that develops foundational AI research and openly released models, datasets, code, and tools. Its work includes Olmo language models, Molmo multimodal models, Tülu instruction-following models, and research in areas such as scientific discovery, robotics, and conservation. Researchers, developers, and other organizations can use its models and datasets to build and study AI systems; Ai2 also provides documentation and a platform for exploring its models.
Ai2 introduces ToyBlocks, an open simulation framework for robot learning
Ai2 introduced ToyBlocks, a simulated tabletop environment and framework for generating robot-learning tasks and demonstrations. It is designed to let researchers vary training experiences and evaluate robot behavior.
Ai2 and Google begin a climate-modeling collaboration with Google.org funding
Ai2 announced a climate-modeling initiative backed by multimillion-dollar, three-year funding from Google.org and technical support from Google Research. The work will extend NeuralGCM toward long-term climate projections.
Ai2 is a Seattle-based nonprofit AI research institute that develops open models and foundational AI research. · Model developers
How they use Nebius AI Cloud
Ai2 participated in TMax, an open research project developing reinforcement learning recipes for terminal agents. The paper identifies work performed while Hamish Ivison and Nathan Lambert were at Ai2 and credits the institute’s Beaker team with infrastructure help.
The authors acknowledge Nebius cloud credits that helped run evaluations for the project. This disclosure establishes research evaluation use. The paper also describes model training and sandbox infrastructure, but does not attribute that training infrastructure to Nebius.
TMax: A simple recipe for terminal agents — second version
Listed 19 Nov 2026 · arXiv · Web page
The TMax authors credit Ai2’s Beaker team with infrastructure help and Nebius cloud credits with supporting research evaluations. The second version’s arXiv publication header is dated 2 November 2026.
TMax: A simple recipe for terminal agents — second versionListed 19 Nov 2026 · arXiv · Web pageThe TMax authors credit Ai2’s Beaker team with infrastructure help and Nebius cloud credits with supporting research evaluations. The second version’s arXiv publication header is dated 2 November 2026.
Ai2 (the Allen Institute for Artificial Intelligence) is a Seattle-based nonprofit research institute that develops foundational AI research and openly released models, datasets, code, and tools. Its work includes Olmo language models, Molmo multimodal models, Tülu instruction-following models, and research in areas such as scientific discovery, robotics, and conservation. Researchers, developers, and other organizations can use its models and datasets to build and study AI systems; Ai2 also provides documentation and a platform for exploring its models.
Ai2 introduces ToyBlocks, an open simulation framework for robot learning
Ai2 introduced ToyBlocks, a simulated tabletop environment and framework for generating robot-learning tasks and demonstrations. It is designed to let researchers vary training experiences and evaluate robot behavior.
Ai2 and Google begin a climate-modeling collaboration with Google.org funding
Ai2 announced a climate-modeling initiative backed by multimillion-dollar, three-year funding from Google.org and technical support from Google Research. The work will extend NeuralGCM toward long-term climate projections.