---
title: "Diffusion Language Model"
description: "A diffusion language model generates text by repeatedly refining a partially specified sequence rather than committing to every next token strictly from left to right. In a masked-token design, generation begins with placeholder positions and iteratively predicts which tokens belong there. Some systems refine an entire response; others generate and refine blocks in sequence."
canonical_url: "https://promptmetheus.com/resources/llm-knowledge-base/diffusion-language-model"
---

> Index of every page on this site: [llms.txt](https://promptmetheus.com/llms.txt)

A diffusion language model generates text by repeatedly refining a partially specified sequence rather than committing to every next [token](https://promptmetheus.com/resources/llm-knowledge-base/token) strictly from left to right. In a masked-token design, generation begins with placeholder positions and iteratively predicts which tokens belong there. Some systems refine an entire response; others generate and refine blocks in sequence.

That differs from the usual autoregressive [Large Language Model (LLM)](https://promptmetheus.com/resources/llm-knowledge-base/large-language-model-llm), which conditions each new token on the preceding tokens and cannot revise an earlier token during the same decode pass. A diffusion model can use information from both sides of a position within the region being refined. The architecture and training objective determine how that refinement works; text diffusion is a family of approaches, not one fixed decoding algorithm.

Parallel updates may improve latency when a model runs at low concurrency on capable hardware. The trade-off is that several refinement passes can still be needed, and a faster single request does not necessarily imply better throughput or lower cost in a heavily batched service. Quality, controllability, streaming behavior, and the number of refinement steps should be evaluated on the intended workload rather than inferred from tokens-per-second alone.

This term is useful when comparing generation architectures and [inference](https://promptmetheus.com/resources/llm-knowledge-base/inference) strategies. It is separate from [model distillation](https://promptmetheus.com/resources/llm-knowledge-base/model-distillation), which transfers behavior between models, and from [Mixture of Experts (MoE)](https://promptmetheus.com/resources/llm-knowledge-base/mixture-of-experts-moe), which controls which parameters are active. A model can combine diffusion-style generation with an MoE backbone.

Google's [DiffusionGemma technical introduction](https://blog.google/innovation-and-ai/technology/developers-tools/diffusion-gemma-faster-text-generation/) describes one blockwise implementation and the workload limits of its reported speed gains.

## Sitemap

See the full [sitemap](https://promptmetheus.com/sitemap.md) for all pages.