--- title: 10x Faster Than Traditional LLMs | daily.dev description: A founder recounts a 2024 research breakthrough in which a diffusion-based transformer model matched the perplexity of a GPT-2 scale autoregressive model while... canonical: https://daily.dev/posts/10x-faster-than-traditional-llms-whrcydvjf twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: 10x Faster Than Traditional LLMs | daily.dev og:description: A founder recounts a 2024 research breakthrough in which a diffusion-based transformer model matched the perplexity of a GPT-2 scale autoregressive model while... og:url: https://daily.dev/posts/10x-faster-than-traditional-llms-whrcydvjf og:image: https://api.daily.dev/og/posts/WHRcydVjf.png og:image:alt: 10x Faster Than Traditional LLMs og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # 10x Faster Than Traditional LLMs **[No Priors: AI, Machine Learning, Tech, & Startups](https://daily.dev/sources/nopriorspodcast)** · 1 min read · 0 upvotes · 0 comments ## Summary A founder recounts a 2024 research breakthrough in which a diffusion-based transformer model matched the perplexity of a GPT-2 scale autoregressive model while generating text roughly 10x faster, since diffusion outputs many tokens in parallel. This result motivated starting a company to scale the approach into commercial diffusion-based language models. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How much faster is a diffusion-based language model compared to a traditional autoregressive LLM at the same quality? A diffusion-based transformer model generated text about 10 times faster than an equivalent autoregressive model, while matching its perplexity at a GPT-2 scale of under a billion parameters. The speedup comes from diffusion models outputting many tokens simultaneously rather than one at a time, as autoregressive models do, while fitting training data just as well. _Developers evaluating inference speed tradeoffs in LLM architectures can follow research like this on daily.dev._ ## Similar posts on daily.dev - [The Anatomy of Diffusion LLMs](https://daily.dev/posts/the-anatomy-of-diffusion-llms-41s7k4vg9) · Daily Dose of Data Science \| Avi Chawla \| Substack · 1 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#diffusion-models](https://daily.dev/tags/diffusion-models) [View this post on daily.dev](https://daily.dev/posts/10x-faster-than-traditional-llms-whrcydvjf) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"10x Faster Than Traditional LLMs","url":"https://daily.dev/posts/10x-faster-than-traditional-llms-whrcydvjf","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/10x-faster-than-traditional-llms-whrcydvjf"},"datePublished":"2026-10-01T13:40:27.183Z","dateModified":"2026-10-01T13:40:42.214Z","description":"A founder recounts a 2024 research breakthrough in which a diffusion-based transformer model matched the perplexity of a GPT-2 scale autoregressive model while...","image":"https://i.ytimg.com/vi/FLdF462oWNY/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/FLdF462oWNY/sddefault.jpg","isAccessibleForFree":true,"articleSection":"No Priors: AI, Machine Learning, Tech, & Startups","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"No Priors: AI, Machine Learning, Tech, & Startups","logo":"https://media.daily.dev/image/upload/s--Hj2kMkm_--/f_auto,q_auto/v1772391934/logos/nopriorspodcast","url":"https://daily.dev/sources/nopriorspodcast"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/10x-faster-than-traditional-llms-whrcydvjf","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,diffusion-models","timeRequired":"PT1M","video":{"@type":"VideoObject","name":"10x Faster Than Traditional LLMs","description":"A founder recounts a 2024 research breakthrough in which a diffusion-based transformer model matched the perplexity of a GPT-2 scale autoregressive model while...","thumbnailUrl":"https://i.ytimg.com/vi/FLdF462oWNY/sddefault.jpg","uploadDate":"2026-10-01T13:40:27.183Z","duration":"PT1M","url":"https://api.daily.dev/r/WHRcydVjf","embedUrl":"https://www.youtube.com/embed/FLdF462oWNY"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"No Priors: AI, Machine Learning, Tech, & Startups","item":"https://daily.dev/sources/nopriorspodcast"},{"@type":"ListItem","position":3,"name":"10x Faster Than Traditional LLMs"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/10x-faster-than-traditional-llms-whrcydvjf#faq","mainEntity":[{"@type":"Question","name":"How much faster is a diffusion-based language model compared to a traditional autoregressive LLM at the same quality?","acceptedAnswer":{"@type":"Answer","text":"A diffusion-based transformer model generated text about 10 times faster than an equivalent autoregressive model, while matching its perplexity at a GPT-2 scale of under a billion parameters. The speedup comes from diffusion models outputting many tokens simultaneously rather than one at a time, as autoregressive models do, while fitting training data just as well. Developers evaluating inference speed tradeoffs in LLM architectures can follow research like this on daily.dev."}}]} ```