thx everyone for coming! had great convos with incredibly inspiring ppl, was a privilege!

Ai2@ai2.bsky.social · 2h ago

📸 140+ people joined us during #SeattleTechWeek to learn how AI gets built at Ai2. Research lead Iz Beltagy walked through what it takes to train a truly open LLM—from large-scale training runs to eval before release. Thanks to everyone who came + asked thoughtful questions. 🤝

really cool use of our open source infini-gram engine to study llm regurgitation

Ai2@ai2.bsky.social · 5d ago

When a model writes, where do its words come from? Are they new, or do they match exactly with language it saw in training? An AI-writing detector can't tell you. @tuhinchakr.bsky.social's group at Stony Brook has been dissecting AI-generated prose with our infini-gram engine. 🧵

learn about the truly impressive infra behind our olmoearth platform 👇

Ai2@ai2.bsky.social · last wk.

The organizations best positioned to use Earth-observation models – those working in conservation, food security, & disaster response – often can't run these models at the scale they need. That's an infrastructure problem. So we built the OlmoEarth Platform to solve it. 🧵

fully open models are important, and we wholeheartedly support their development to advance science for everyone 🚀

Ai2@ai2.bsky.social · 2w ago

As a nonprofit research institute dedicated to advancing open science, we're encouraged to see growing support for open models across the AI ecosystem. We believe the evidence behind advanced AI systems shouldn’t be locked up in a few hands.

our incredible completely free service gets even better 🚀

Ai2@ai2.bsky.social · 2w ago

Two updates to Asta, our ecosystem of AI agents for science: a one-click handoff from AutoDiscovery to Asta’s data analysis tools, & paper search that evaluates its own results + searches again when they fall short. 🧵

i'll be in seattle for this—come say hi!

Ai2@ai2.bsky.social · 3w ago

What does it actually take to build cutting-edge AI systems? On July 30 during #SeattleTechWeek, the researchers behind Ai2's open models sit down to talk through the deep technical work behind them. 🧵

We built SciArena to test how well AI models handle scientific literature questions, as judged by researchers. It's retiring July 15, and the results are in: ~1,700 users cast ~3,900 votes. Here's what they told us. 🧵

Our @skylightmarine.bsky.social team is rolling out Shippy, an AI agent for the people protecting our ocean in real time. A wrong answer at these stakes can send a patrol vessel miles off course. Here's the architecture we built so analysts can trust it. 🧵

olmOCR 2 is now in the Ai2 Playground—our home for our fully open text, video, and image understanding models. 🧵

The Danish Foundation Models (DFM) project is adapting our modular FlexOlmo architecture into a lighter-weight system that runs on commodity hardware—putting collaborative model building within reach of smaller research groups & organizations. 🧵

Today we're releasing OlmoEarth v1.2, the latest in our family of open foundation models for Earth observation. 🌍 We've switched to rotary positional embeddings (RoPE), which reduces artifacts in the embeddings & gives a small performance boost. 🧵

Hybrid (transformer–RNN) models are fast becoming a serious alternative to the transformer, but a big question remains: how do they process tokens differently & how does this impact performance? We compared our transformer (Olmo 3) & hybrid (Olmo Hybrid) models to find out. 🧵

Learn how @thinkaisquared.bsky.social & Domyn used Olmo, our family of fully open language models, to build their own models for regulated industries like finance, healthcare, & the public sector. 🧵

We're releasing MolmoMotion, a 3D motion forecasting model. Given one or a few video frames, 3D points on an object, & an instruction like "Put the white bowl on the table," MolmoMotion predicts where those points will go over the next few seconds in a shared 3D world frame. 🧵

Building an LLM means evaluating it over & over as it changes. Tweak a hyperparameter or scale the model up, & every new checkpoint sends you back through the same benchmarking loop. We're releasing olmo-eval, a workbench built for this kind of iterative model development. 🧵

LLMs are no longer created w/ human data alone. They rely on other models to generate & filter data, evaluate outputs, & guide dev work. So what is a modern LLM built on? Olmo 3 → 89 model + 183 dataset dependencies; Nemotron 3 → 273 + 560 We made ModSleuth to trace this. 🧵

𝗔𝗖𝗘𝟮𝗦-𝗦𝗛𝗶𝗘𝗟𝗗+, our new climate emulator that learns to separate the effects of sea surface temperature & CO2, is now on @hf.co—check it out → huggingface.co/allenai/ACE2...

allenai/ACE2S-SHiELD-plus · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

huggingface.co

Ai2@ai2.bsky.social · 2mo ago

Today we're introducing ACE2S-SHiELD+, a climate emulator that learns to separate the effects of sea surface temperature & CO2. It accurately handles scenarios where previous versions of our ACE family of climate emulators produced inaccurate results. 🧵

Today we’re releasing OlmoEarth v1.1. It’s 3x cheaper to run than v1 while delivering the same state-of-the-art performance—and fully open. 🧵

Brendan Works is a product manager focused on paratransit services in Seattle. See how he built PointCheck, a website accessibility checker powered by our open Molmo, MolmoWeb, & Olmo 3 models. 👇

Now available in AstaLabs in limited research preview: MyScholarQA, a personalized version of ScholarQA for scientific deep research. ScholarQA helps synthesize evidence from 12M+ open-access papers. MyScholarQA adds user profiles to tailor that synthesis to you. 🧵

Artificial Analysis relies on our IFBench eval to test how closely models follow user prompts. Most evals in their Intelligence Index saturate within months. IFBench hasn't because it measures what others miss—and what frontier models still struggle with. 🧵

Today we’re releasing EMO, a new mixture-of-experts (MoE) model trained so modular structure emerges directly from data without human-defined priors. EMO can use a small subset of its experts for a given task while keeping near full-model performance. 🧵

Robotics models often struggle outside controlled environments. Ours is built to work in real ones. Today we're launching MolmoAct 2, which can assist with a host of chores & lab tasks, plus the MolmoAct 2-Bimanual YAM dataset—the largest open robotics dataset of its kind. 🧵

Today we published a Q&A with Interim CEO Peter Clark on what’s next for Ai2, from advancing truly open AI systems to applying AI in areas like scientific discovery & the planet. 👇

New AstaBench results show frontier models making progress on scientific research, but the benchmark remains far from solved. Claude Opus 4.7 leads overall at 58.0%, while GPT-5.5 comes within 5.1 points at less than half the measured cost per problem. 🧵

Last year, we introduced FlexOlmo, a novel way to train parts of a model independently then combine them later. BAR builds on that idea for a harder problem: how to keep improving a model without having to retrain each time. 🧵

You can now train, adapt, and eval web agents on your own tasks. We're releasing the full MolmoWeb codebase—the training code, eval harness, annotation tooling, synthetic data pipeline, & client-side code for our demo. 🧵

Today we're releasing WildDet3D—an open model for monocular 3D object detection in the wild. It works with text, clicks, or 2D boxes, and on zero-shot evals it nearly doubles the best prior scores. 🧵

MolmoBot, our open robotic manipulation suite trained entirely in simulation, now has code, training data, a data generation pipeline, & evals all available. This puts our robotics models within reach of any research lab—no extensive real-world data collection required. 🧵