🔍 Read the full analysis: What Is AstaBrief? A Look At Asta’s Report-Generation Model on ThorstenMeyerAI.com
Get monitors, keyboards and dev gear delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
TL;DR
Ai2 has released AstaBrief 8B, an open-weights model for generating reports from a research question and retrieved scientific literature. Ai2 reports an average generation time of 51.1 seconds in Asta’s Fast mode, versus 178.5 seconds for its Claude-powered Thinking mode, but has not rerun its full evaluation against current frontier models.
Ai2 has released AstaBrief 8B, an open-weights model designed to produce cited scientific reports from a research question and retrieved literature excerpts, as described in the original analysis. The model is available as Fast mode in Ai2’s Asta platform, and the institute says its full pipeline averaged 51.1 seconds per report, compared with 178.5 seconds for Asta’s Claude-powered Thinking mode.
Ai2 says AstaBrief is built on Qwen3-8B and adapted for long-form scientific synthesis. Given a query and relevant retrieved snippets, it generates a report in one pass. According to the institute, this bypasses intermediate snippet-summarization and clustering steps used in Thinking mode, as well as writing the report section by section.
The release includes the model weights, training data and an example workflow that researchers can adapt, including for reports based on their own PDFs. Ai2 says the model was trained with supervised fine-tuning and direct preference optimization. The team focused on creating and filtering examples that teach report-writing behavior, including attention to citation grounding, rather than using the reinforcement-learning approach it considered.
Ai2’s timing figures cover the full Asta pipeline, not just the model’s text generation. The reported averages make Fast mode about 3.5 times faster than Thinking mode, based on the numbers provided. They describe speed, however, and do not by themselves establish that the reports are equally accurate, complete or well-supported by their citations.
Faster Reports, Open Deployment Options
The release gives research groups a model they can inspect and adapt for a common task: comparing findings across scientific literature while applying constraints such as a method, population or setting. AstaBrief’s reported speed could make it easier to generate and revisit reports as working research documents, rather than waiting several minutes for each run.
Open weights and a sample workflow also offer a route for institutions to experiment with local deployment. Ai2 says this may suit work involving sensitive questions or unpublished research, where teams may prefer to run tools on their own infrastructure. The release does not establish that every institution can deploy it without additional technical work, or that a local setup will match the service’s reported speed.
For scientific use, the central test is not response time alone. A useful synthesis must represent the limits of the underlying studies, avoid leaving out relevant evidence and provide citations that actually support its statements. Those questions matter whether a report is produced in under a minute or several minutes.
scientific report writing software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How Asta’s Two Modes Differ
Asta is Ai2’s platform for scientific work, and its report-generation feature now offers AstaBrief Fast mode alongside Thinking mode. Ai2 describes Thinking mode as Claude-powered. Its supplied description says that mode uses additional stages to summarize and cluster retrieved snippets before drafting a report section by section, whereas AstaBrief takes retrieved excerpts and produces a report in one pass.
Ai2 says its training and evaluation work used real research queries and citation-focused filtering, with comparisons to proprietary models that reflected the frontier at the time. The institute says most of this work was completed in 2025. That timing limits what the reported comparisons can show about how AstaBrief stacks up against models available now.
The open release is part of Ai2’s broader work on adapting open models to scientific needs. The institute also points to work with scientific communities through the NSF OMAI initiative and says it expects to share further findings. That broader effort provides context for the release, but it does not substitute for independent testing of AstaBrief’s report quality.
““We wanted to help scientists generate cited reports faster, with a model they could download and run themselves.””
— Ai2
As an affiliate, we earn on qualifying purchases.
Report Quality Still Needs Testing
Ai2 has not rerun its full evaluation against current frontier models, and the announcement does not establish a present-day head-to-head comparison of report quality. The reported timing averages are specific, but the material does not provide enough detail to independently verify the claimed quality comparison between Fast and Thinking modes.
Other details remain open: how Ai2 measured report quality; how often citations accurately support the claims they accompany; and how results vary by scientific field or query type. The supplied information also does not specify the hardware and configuration behind the timing averages or establish whether a locally run model performs identically to Asta’s hosted Fast mode. Independent evaluations would be needed to answer those questions.
citation management tools for research
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Independent Tests Will Add Evidence
Researchers can examine the released weights and training data and adapt Ai2’s example workflow to their own literature or PDFs. Tests across disciplines and deployment settings could show where the model produces useful syntheses, where it misses evidence and whether its citations remain verifiable.
Ai2 says it expects to share further results from its broader work on adapting open models for scientific use. A new full evaluation against current frontier models has not yet been reported. Until then, the clearest confirmed comparison is Ai2’s stated pipeline timing, not a current independent judgment that the faster mode matches or exceeds other models in report quality.
As an affiliate, we earn on qualifying purchases.
Key Questions
What does AstaBrief 8B do?
AstaBrief 8B generates cited scientific reports from a research question and literature excerpts retrieved for that query. Ai2 describes its report generation as a one-pass process.
How fast is AstaBrief compared with Thinking mode?
Ai2 reports an average of 51.1 seconds per report for Fast mode across Asta’s full pipeline, compared with 178.5 seconds for its Claude-powered Thinking mode. These are Ai2’s figures; the announcement does not specify the hardware and configuration behind the averages.
Is AstaBrief open source?
Ai2 describes the release as open-weights and says it includes model weights, training data and an example workflow. The supplied material does not detail every term governing use or redistribution, so users should check the release documentation for licensing details.
Has Ai2 shown that AstaBrief matches current frontier models?
No current full comparison is reported. Ai2 says most of its model work was completed in 2025 and that it has not rerun the full evaluation against today’s frontier models. The reported speed figures do not establish comparative report quality.
Can researchers run AstaBrief locally?
Ai2 says open weights could let institutions run the model on their own infrastructure and provides an example workflow researchers can adapt, including for reports from their own PDFs. The supplied information does not show whether local performance matches Asta’s hosted Fast mode.
Primary source: Hugging Face · via ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
