DiffuMamba: High-Throughput Diffusion LMs with Mamba Backbone

TL;DR

DiffuMamba introduces a new diffusion language model with a Mamba backbone that significantly improves inference throughput and efficiency for long sequences, matching performance of Transformer-based models.

Contribution

The paper presents DiffuMamba, a novel diffusion language model with a Mamba backbone, combining diffusion objectives with linear-time sequence modeling, and demonstrates its superior efficiency and competitive performance.

Findings

DiffuMamba achieves up to 8.2x higher inference throughput.

Models match Transformer-based diffusion in downstream tasks.

Cache-efficient block diffusion with Mamba mixers scales linearly with sequence length.

Abstract

Diffusion language models (DLMs) have emerged as a promising alternative to autoregressive (AR) generation, yet their reliance on Transformer backbones limits inference efficiency due to quadratic attention or KV-cache overhead. We introduce DiffuMamba, a masked diffusion language model built on a bidirectional Mamba backbone that combines the diffusion objective with linear-time sequence modeling, and DiffuMamba-H, a hybrid variant with interleaved attention. Across scales up to 1.3B parameters, our models match Transformer-based diffusion in downstream performance while achieving up to 8.2x and 4.3x higher inference throughput, respectively, on long sequences. We further present a systematic analysis of inference efficiency across modern DLM variants combining asymptotic complexity with empirical measurements. Notably, cache-efficient block diffusion with Mamba mixers emerges as the only strategy that scales linearly with sequence length and achieves the strongest performance across all baselines, suggesting a promising direction for future diffusion-based generation systems.