DiffThinker: Towards Generative Multimodal Reasoning with Diffusion Models

TL;DR

DiffThinker introduces a diffusion-based generative framework for multimodal reasoning, reformulating it as an image-to-image task to improve logical consistency and spatial accuracy in vision-centric applications.

Contribution

The paper presents DiffThinker, a novel diffusion-based paradigm for multimodal reasoning, demonstrating significant performance improvements over existing models in vision-centric tasks.

Findings

Outperforms GPT-5 by +314.2% in certain tasks

Achieves +111.6% improvement over Gemini-3-Flash

Surpasses fine-tuned Qwen3-VL-32B baseline by +39.0%

Abstract

While recent Multimodal Large Language Models (MLLMs) have attained significant strides in multimodal reasoning, their reasoning processes remain predominantly text-centric, leading to suboptimal performance in complex long-horizon, vision-centric tasks. In this paper, we establish a novel Generative Multimodal Reasoning paradigm and introduce DiffThinker, a diffusion-based reasoning framework. Conceptually, DiffThinker reformulates multimodal reasoning as a native generative image-to-image task, achieving superior logical consistency and spatial precision in vision-centric tasks. We perform a systematic comparison between DiffThinker and MLLMs, providing the first in-depth investigation into the intrinsic characteristics of this paradigm, revealing four core properties: efficiency, controllability, native parallelism, and collaboration. Extensive experiments across four domains (sequential planning, combinatorial optimization, constraint satisfaction, and spatial configuration) demonstrate that DiffThinker significantly outperforms leading closed source models including GPT-5 (+314.2\%) and Gemini-3-Flash (+111.6\%), as well as the fine-tuned Qwen3-VL-32B baseline (+39.0\%), highlighting generative multimodal reasoning as a promising approach for vision-centric reasoning.