Theme
AI Visuals
OUI-1
OUI-1 is a specialized text-diffusion model from Thesys for generating interface screens as OpenUI Lang code.
Give it the component signatures from a UI library and a plain-language screen brief, and it returns a structured screen that an OpenUI renderer can validate and display. Use this as a first read, not a recommendation. Open the original project before trusting details like terms, limits, privacy, cost, setup, or safety.
What it is
A model trained for interface screens
OUI-1 is a fine-tune of Google's DiffusionGemma 26B-A4B-it with 26 billion total parameters and 4 billion active parameters. Its job is narrow: write OpenUI Lang programs for user-interface screens rather than act as a general chatbot.
Why it stands out
Generates blocks instead of one token at a time
The text-diffusion design works on 256-token blocks. Thesys reports about one second for a light screen and three to six seconds for a dense one in its single-request A100 80GB benchmark setup.
Availability
Public weights, specialist setup
The merged BF16 weights and LoRA adapter are published on Hugging Face under the Gemma Terms of Use. The documented paths use vLLM 0.24 or newer, Transformers 5.11 or newer, and OpenUI packages for prompt generation, validation, and rendering.
Why it matters
What makes it useful
OUI-1 is trained around one compact UI language and the components an application explicitly allows. That gives builders a focused, inspectable model aimed at the latency-sensitive step between a request and a rendered screen.
What to know
Where it fits
It fits OpenUI-based assistants and applications that need to turn a request or tool result into a structured interface. A builder supplies the component contract, validates the returned program, and renders it through React, Vue, Svelte, or another OpenUI integration.
Notable points
What stands out
Thesys reports 132 valid outputs from 184 attempts, or 71.7%, on its Generative UI Benchmark, compared with 24 of 184 for the base model. The benchmark repository publishes the briefs, validator, raw outputs, and scored results so the runs can be rescored offline.
Before using
What to review
The recommended FP8 vLLM setup uses about 25.8 GiB for the weights; the documented BF16 Transformers path peaks at roughly 52 GiB of GPU memory.
The model generates OpenUI Lang, not a finished product. Validate its output and let the component library constrain props, data, interactions, and rendering.
Tool calling works through the Gemma 4 format, but the model card notes that required or named tool choices are not enforced by the current vLLM path.
The checkpoint is a Gemma derivative. Check the current Gemma Terms of Use from the model card before adopting or distributing it.
Reader fit
Who may find it relevant
Builders experimenting with generative interfaces inside assistants, copilots, dashboards, forms, or agent applications.
Teams that already have a component library and want model output limited to those registered building blocks.
Less relevant for people seeking a general chat model, a no-code visual editor, or a lightweight model for an ordinary consumer laptop.
Editorial note
Why LifeHubber lists it
LifeHubber lists OUI-1 because it makes the model layer of generative UI unusually concrete: the weights, prompt contract, output language, benchmark data, and rendering path are all inspectable. Readers can decide whether that focused speed and control are worth the substantial GPU requirement and an OpenUI-specific workflow.
Source links
Source materials
Reader note
Before relying on this entry
LifeHubber lists entries to help readers inspect AI projects, not to endorse them or prove they are safe, suitable, accurate, maintained, or right for a specific use. We do not verify every entry in depth. Before relying on anything listed, review the original materials, terms, privacy practices, limits, and risks that matter for your situation.
What to explore next
Compare the model with a broader generative UI stack.
OUI-1 generates OpenUI screens from a component contract. Continue with an application framework that also covers chat, shared state, tool rendering, and human input around generated interfaces.
More in AI Models
Keep browsing this category
Explore more AI model resources.
Gemma 4
google/gemma-4
A Google DeepMind Gemma 4 model family collection with public checkpoints including Gemma 4 12B, a dense multimodal model Google describes around local agentic workflows, native audio input, and encoder-free vision/audio handling.
MiniCPM5-2B
OpenBMB/MiniCPM5-2B
An OpenBMB 2.52B-parameter language model for local assistants, coding agents, tool use, reasoning, and long-context workflows, with 131K context plus BF16, GGUF, MLX, GPTQ, and DSpark variants.
Ling 3.0
inclusionAI/ling-30
An inclusionAI sparse language-model collection led by Ling-3.0-flash, with standard and FP8 checkpoints for coding, tool use, long-context work, and efficiency-focused agent serving.
For project maintainers
Listed here? You can use the badge.
If you maintain a project with a current LifeHubber listing, you may add the optional “Listed on LifeHubber AI Resources” badge to its README, docs, or website. No introduction or permission request is needed.
