# DiffResearch framework uses diffusion models to cut research AI latency

> The open‑source DiffResearch system swaps a diffusion language model for a traditional autoregressive one, achieving about six‑fold faster end‑to‑end speed on benchmark tasks.

Oossa · 2026-10-22 · https://oossa.com/en/diffresearch-framework-uses-diffusion-models-to-cut-research-ai-latency

DiffResearch is a new multi‑agent research framework that puts a diffusion language model (dLLM) at the core of planning, retrieval and writing. The authors compared the diffusion backbone with a similar autoregressive model on 100 Deep Research Bench tasks. The diffusion version ran roughly six times faster overall, mainly because planning and long‑form writing became much quicker. A special mode called dFast reads intermediate denoising states and splits the work into subqueries, giving an additional median 1.99× speed boost without hurting report quality. The code is released under an Apache‑2.0 license.

## The facts

- Diffusion backbone is about 6× faster end‑to‑end on the benchmark.
- dFast mode adds a median 1.99× speedup for decomposition.

## Why it matters

Researchers can get AI‑generated reports in minutes instead of tens of minutes, making interactive deep‑research tools more practical.

## Sources & references

1. [DiffResearch: Accelerating Deep Research with Diffusion Language Models](https://arxiv.org/abs/2610.28909) – arXiv, 2026-10-22

Last updated: 2026-10-22