---
title: "Open-sourcing AstaBrief, the fast report-generation model in Asta"
description: "AstaBrief is an open-weights 8B-parameter model developed by Ai2 for generating cited scientific reports from a research question and literature excerpts, designed for evidence-grounded outputs. It was trained from Qwen3-8B using supervised fine-tuning on 47K filtered real user queries and direct preference optimization, focusing on data quality. Unlike Asta’s Thinking mode, which uses a multi-step pipeline, AstaBrief writes the full report in a single pass to reduce generation time. In Asta’s Fast mode, it averages 51.1 seconds per report versus 178.5 seconds for Thinking mode, roughly 3.5× faster, while competitive on answer and citation quality. The model and training data are open-sourced on Hugging Face, enabling institutions to deploy behind their own firewall and supporting future open scientific models."
---

# Open-sourcing AstaBrief, the fast report-generation model in Asta

[Hugging Face](https://yomu.fyi/company/hugging-face) · Kyle Wiggers · Oct 2, 2026

**Type:** Announcement

## Summary

AstaBrief is an open-weights 8B-parameter model developed by Ai2 for generating cited scientific reports from a research question and literature excerpts, designed for evidence-grounded outputs. It was trained from Qwen3-8B using supervised fine-tuning on 47K filtered real user queries and direct preference optimization, focusing on data quality. Unlike Asta’s Thinking mode, which uses a multi-step pipeline, AstaBrief writes the full report in a single pass to reduce generation time. In Asta’s Fast mode, it averages 51.1 seconds per report versus 178.5 seconds for Thinking mode, roughly 3.5× faster, while competitive on answer and citation quality. The model and training data are open-sourced on Hugging Face, enabling institutions to deploy behind their own firewall and supporting future open scientific models.

## Context

Asta is an agentic platform for scientific work where users generate cited reports from literature. The post describes the development of AstaBrief, an open-weights model for faster report generation, and its integration as Fast mode alongside the existing Thinking mode.

## Approach / What changed

AstaBrief was trained from Qwen3-8B using supervised fine-tuning on 47K filtered real user queries and direct preference optimization on pairwise preference data. It generates the full report in a single pass, bypassing the multi-step snippet summarization and section-by-section writing used by the Thinking mode. The model is deployed as Fast mode in Asta and open-sourced on Hugging Face.

## Takeaways

- AstaBrief generates reports in a single pass, achieving roughly 3.5× faster generation than the multi-step Thinking mode (51.1 seconds vs 178.5 seconds average).
- The model was trained with a focus on data quality over complex reinforcement learning, using SFT and DPO on real user queries and filtered training data.
- Open-sourcing allows institutions to deploy AstaBrief on their own infrastructure behind a firewall, addressing privacy concerns for sensitive research.

## FAQ

### How does AstaBrief achieve faster report generation?

It writes the full report in a single pass instead of using the multi-step pipeline with snippet summarization and section-by-section writing, reducing generation time.

### What training data was used for AstaBrief?

It used 47K filtered examples from real user queries on Asta, with target reports generated by proprietary models like Claude and GPT-4.1, and DPO pairs from separate queries.

### Why was AstaBrief open-sourced?

To allow institutions to run the model on their own hardware behind a firewall for sensitive research, and to enable the research community to study and build on the approach.

**Tags:** [AI](https://yomu.fyi/topic/ai), [Fine-Tuning](https://yomu.fyi/topic/fine-tuning), [LLMs](https://yomu.fyi/topic/llm), [Open Source](https://yomu.fyi/topic/open-source), [RAG](https://yomu.fyi/topic/rag)

- Source: [Hugging Face](https://huggingface.co/blog/allenai/astabrief)
- Source URL: https://huggingface.co/blog/allenai/astabrief
- Ingested by Yomu: 2026-10-02T18:00:11.351Z

[Read original post](https://huggingface.co/blog/allenai/astabrief)