Ai2 open-sources AstaBrief scientific report-generation model
Ai2 has released AstaBrief 8B, an open-weights model designed to turn a research question and retrieved literature excerpts into a cited scientific report. It is available in Asta’s Generate a report feature as Fast mode, alongside the Claude-powered Thinking mode. Ai2 is also publishing the training data and an example workflow for generating reports from users’ own PDFs.
The model was built from Qwen3-8B and trained mainly with supervised fine-tuning and direct preference optimization. Its data pipeline began with 90K research-focused queries; quality filtering produced 47K supervised examples and about 6K preference examples. AstaBrief generates a complete report in one pass, avoiding several stages used by the Thinking mode.
Across the full Asta pipeline, Fast mode averaged 51.1 seconds per report, compared with 178.5 seconds for Thinking mode. Ai2 said the open weights could allow institutions to run the model on their own infrastructure, including for sensitive or unpublished research. Most training and evaluation was completed in 2025, and the published comparisons were not rerun against current frontier models.
Written by Sakhanda Wire AI from reporting by Hugging Face. Facts, quotes and full context are in the original article below.