Ai2 Newsletter
October 2025
Top story - Asta DataVoyager: Data-driven discovery and analysis
Scientists don't lack for questions they want answered—they lack hours in the day and tools they can trust to get those answers. Experimental logs live in spreadsheets; instrument readings arrive as CSVs; and results tables pile up across projects. Turning those structured files into answers takes time and often requires advanced programming skills to be done efficiently.
To fill the gap, we’re launching DataVoyager in Asta, our ecosystem for scientific research agents. Built to address the challenges scientists face in drilling down into structured datasets, Asta DataVoyager delivers data-driven discovery and analysis capabilities that allow you to ask questions about structured files in plain language and get clearly cited, explainable answers with copyable code, clear visuals, and a concise, well-supported summary.
Asta DataVoyager was designed from the start to be intuitive for users, regardless of their comfort level working with dataset analysis tooling. Users upload a dataset and ask a question (e.g., “Which treatment shows the most improvement after week 6?”), along with an optional prompt to establish context so that Asta DataVoyager makes better initial choices.
The Cancer AI Alliance (CAIA) has already prototyped a federated instance of Asta DataVoyager on its platform. CAIA – which unites four leading cancer centers – earlier this week announced its federated learning platform that enables AI models to travel to each participating cancer center’s secure data store to learn from it locally, generating a summary of its learnings without individual clinical data ever leaving institutional firewalls. Ai2’s engagement with CAIA is made possible by generous support from Allen Family Philanthropies.
Read the CAIA announcement here.
"We are excited about the possibility of providing powerful and secure analytics tools to cancer researchers who may not have AI expertise," says Jeff Leek, PhD, VP and Chief Data Officer at Fred Hutch Cancer Center and Holder of the J. Orin Edson Foundation Endowed Chair. Leek is also the founder and scientific director of CAIA. "When I think about the future of where I want it to go, I think about this tool in the hands of clinicians, helping to answer important questions that will ensure the best possible care for cancer patients.”
Importantly, Asta DataVoyager's output is structured and largely consistent across runs. That makes it easier to share with collaborators, copy to a lab notebook, or include in a preprint’s supplementary materials without much hand-reformatting—shortening the path from a question to a well-supported answer and making each step visible so you can understand and reliably trust the result.
Asta DataVoyager is a trusted AI collaborator—one that lets researchers make queries about data in natural language and get transparent, reproducible answers they can act on. Moreover, Asta DataVoyager allows teams to stay in full control of their data—they can delete datasets at any time from Asta’s hosted portal or secure on-premises, datacenter, and private cloud deployments.
Reach out to the Asta team to discuss secure deployments and pilot projects, and sign up for updates here.
Asta DataVoyager is already helping drive science forward in early pilots and collaborations. We look forward to hearing what the new feature enables for you—and suggestions to make it even better.
Fluid LLM benchmarking
We introduced Fluid Benchmarking, a more efficient way to test AI models that tailors benchmark evaluation items (e.g., multiple-choice questions) to each model’s skill level instead of using a one-size-fits-all approach. In studies, it delivered more trustworthy scores with ~50× fewer items on some benchmarks.
Build your own AskOlmo bot
We open-sourced the code behind AskOlmo, the chatbot powered by our Olmo model family that's currently live in our Discord server. The repo walks through the setup, wiring real-time chat to Olmo, adding simple commands, and tips for running it on your own server.
ACE2 ML model
In a project with the UK Met Office, our ACE2 weather model showed skill in seasonal forecasting while using far less compute than traditional physics-based systems. ACE2 has certain limitations, but it predicted key climate drivers like the North Atlantic Oscillation and reached correlation scores on par with state-of-the-art physics models.
Compare models in the Ai2 Playground
New in the Ai2 Playground: side-by-side model comparison. Click “Compare models” in the sidebar, choose two of our models, enter a prompt, and view outputs next to each other so the differences stand out. This view helps you judge model quality, sanity-check prompts, and iterate faster.
Ai2 at IUCN World Conservation Congress
Our CEO, Ali Farhadi, will be giving a keynote presentation at the IUCN World Conservation Congress in Abu Dhabi on October 11. Stay tuned for an important announcement related to our work in AI to tackle planet-scale challenges.
AI Innovation in the Open at Seattle AI Week
As part of Seattle AI Week, we're hosting AI Innovation in the Open on October 30—an afternoon of live demos and hands-on tutorials at Ai2 HQ. We’ll kick off with a presentation of our latest research, then you can choose a track: set up and run our upcoming Asta data-driven discovery agent on your own laptop or learn how to customize our Olmo model family using open-source tools.