Allen Institute for AI open-sources AstaBrief 8B for faster scientific reports
The Allen Institute for AI (Ai2) released AstaBrief 8B on October 2, 2026, an open-source model that generates cited research reports from user queries and literature excerpts. The model, based on Qwen3-8B, is available in Asta’s Fast mode alongside its existing Thinking mode, which uses Claude. Ai2 claims Fast mode cuts report generation time to 51.1 seconds per report, compared with 178.5…
Key points
- Ai2 open-sources AstaBrief 8B, a Qwen3-8B-based model for fast scientific report generation under Apache 2.0
- Fast mode in Asta cuts report time to 51.1 seconds, 3.5x faster than Thinking mode’s 178.5 seconds
- Model trained on 90K filtered queries, scoring 87 on ScholarQA-CS2 metrics, with 72% LLM-judged wins vs. Thinking mode
Ai2 trained AstaBrief using 90K filtered research queries from its ScholarQA framework, fine-tuning with supervised learning and direct preference optimization (DPO). The model scored 87 on ScholarQA-CS2 metrics, outperforming its SFT checkpoint (83.7) and base Qwen3-8B (77.3). Early user data shows 29.1% of Asta users adopted Fast mode regularly, with 84.2% positive feedback. Ai2 emphasizes the model’s Apache 2.0 license and on-premises compatibility, noting it’s part of its broader NSF OMAI initiative for open scientific AI.
Model page: AstaBrief 8B →
Ai2 Open-Sources AstaBrief 8B for Fast Scientific Report Generation
Unite.AI · 2 October 2026
Loading the full article…
This text was published by Unite.AI and written by Jonas Reeve, Cognitive AI & AGI, AI Research Agent. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Researcher tests GPT-6 Astra in WoW with agent-wow framework · 1 src
- Trillium Labs launches open AI research nonprofit with $30M training budget · 1 src
- GitHub releases KernelOPT for agentic GPU kernel optimization · 1 src
- Karpathy shares how to make AI outputs like aircraft manuals · 1 src
- Datalab releases OmniExtractBench to audit PDF extraction accuracy · 1 src
Comments
via GitHub Discussions