Top news
Investing in the public benefit OfficialAnthropic NewsOct 18Popular Musk Pledges SpaceX Shares to ‘Every Kid in America’ WebBloomberg TechnologyOct 19Popular Websites are closing their doors to AI agents WebThe Verge AIOct 17 This is the new ChatGPT Voice, powered by GPT Live OfficialOpenAI on YouTubeOct 18 How AI giants will reshape global philanthropy WebAxios AIOct 19 Gemini 4 Argon Really Surprised Me WebTwo Minute PapersOct 19 Google Gemini 4 agents just beat Minecraft with no human intervention WebEngadget AIOct 19 Who could it be? 🧐 OfficialGoogle on YouTubeOct 18 Anthropic gives outside AI watchdog “employee-like” access WebArs Technica AIOct 18 What are self-replicating ‘AI worms’, and why is Anthropic worried about them? WebThe Guardian AIOct 18
GlossaryModelsIntermediate

Diffusion models

Also written: Diffusion, Latent diffusion model, AI diffusion, Diffusion process, Masked Diffusion Language Agents, MDLAs

Generative AI models that create data by reversing a process of gradually adding noise.

In plain terms

Diffusion models are a type of AI that creates new content, like images or videos, by starting with random noise and then slowly refining it. Think of it like a sculptor starting with a block of marble and chipping away until a statue emerges. The AI learns to "un-noise" the random static into a coherent piece of data.

For example, OpenAI's DALL-E uses diffusion models to turn text descriptions into unique images. You might type "a cat wearing a hat" and the model, using a diffusion process, generates a picture matching that description.

The expert version

Diffusion models are a class of generative models that learn to synthesize data by reversing a diffusion process. This process involves gradually adding noise to training data until it becomes pure noise, and the model is trained to denoise this data step-by-step, effectively learning the reverse of the noise-adding process. This iterative denoising allows for the generation of high-fidelity data, though it can be computationally intensive during inference compared to other generative methods.

Why it matters now

Diffusion models are a core generative AI technology, underpinning recent advances in image and video creation. New research is exploring their application beyond media, such as in complex reasoning tasks for AI agents, and developing faster, more efficient variations.

Where it shows up 13 items, newest first
How each company uses it
Anthropic Claude
Anthropic is using diffusion models in new AI agents for complex reasoning, like in chemical analysis.
Google Gemini
Google uses diffusion model concepts in AI features that allow users to re-frame photos after they are taken.
OpenAI ChatGPT
OpenAI has developed and researched diffusion models for image generation, and is working on faster alternatives like consistency models.
Related terms

Generative Adversarial Networks (GANs)Variational Autoencoders (VAEs)Consistency ModelsLatent Diffusion ModelsText-to-Image GenerationDenoising Diffusion Probabilistic Models

Often confused with
Generative Adversarial Networks (GANs)
GANs use two networks, a generator and a
Variational Autoencoders (VAEs)
VAEs learn a compressed representation of data
Autoregressive Models
These models generate data sequentially, predicting the n
Origin

The core ideas behind diffusion models trace back to statistical physics and thermodynamics, with significant contributions from machine learning researchers applying these concepts to generative modeling. Source ↗

Try this about 5 minutes

Use an AI image generator like OpenAI's DALL-E 3. Type a descriptive prompt, such as 'a watercolor painting of a futuristic city at sunset.' Observe the generated image, noting how the AI interprets your text to create a visual output. Experiment with different prompts to see the range of images the model can produce.

Try it: Playbook entries that use thisPlaybook →

The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.

Questions people ask

How do diffusion models create images?
They start with random noise and iteratively remove the noise, guided by learned patterns, until a clear image emerges.
Are diffusion models fast?
Generating images can take time due to the many steps involved, but researchers are developing faster methods and variations.
What kind of data can diffusion models generate?
They can generate various types of data, including images, audio, and even video, by learning from large datasets.

First seen on Subvolts Apr 13, 2022. Updated Oct 16, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.