Skip to content

blog: we taught Google's diffusion model our UI language - #1116

Open
Aditya-thesys wants to merge 1 commit into
mainfrom
blog/diffusiongemma-selfdistill
Open

blog: we taught Google's diffusion model our UI language#1116
Aditya-thesys wants to merge 1 commit into
mainfrom
blog/diffusiongemma-selfdistill

Conversation

@Aditya-thesys

Copy link
Copy Markdown
Contributor

New post: doubling-diffusiongemma. The DiffusionGemma 26B-A4B finetune story: 13% to 57.1% on the public GenUI benchmark, past Gemma 4 31B, at 1.89s per screen on one A100, with the second half of the gain coming from a self-distillation loop.

Contents, blog code only:

  • docs/content/blog/doubling-diffusiongemma.mdx, the post
  • docs/components/charts/diffusion-charts.tsx, six charts in the house chart style
  • docs/lib/diffusion-blog-data.ts, single source of truth for every number
  • docs/public/blog/dg-reveal.html, the animated one-brief-two-models embed
  • docs/mdx-components.tsx, registers the chart components

No other changes.

@vercel

vercel Bot commented Sep 1, 2026

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated
openui-docs Ready Ready Preview Sep 1, 2026 3:20pm UTC

Request Review

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant