AI Frontend Evaluation Suite

Evaluating AI Frontend Output & Harness Wrappers

An open benchmark measuring how 16 AI models and 9 agent harnesses generate real-world web UI. Compare visual outputs, code quality, and accessibility scores side by side.

Prompts Coverage Matrix

Covered Providers (15)

Alibaba DeepSeek Google Meituan Meta MiniMax Moonshot AI OpenAI Poolside StepFun Tencent Thinking Machines Xiaomi Zhipu AI xAI

Benchmark Prompts (8)

Select a prompt to view model output iframe previews and accessibility scores.

Interactive Expert

Digital Garden

Calm, dreamy interactive garden — plant and water virtual plants with soft generative particles, staggered growth animations, and satisfying micro-interactions.

#Creative#Generative#Particles#Interaction#Canvas

Evaluated 16 / 16

Pass Rate 15 / 16 (94%)

Avg Accessibility 89%

Score Range 0% – 98%

Dashboard Expert

High-Density Crypto Trading Terminal

Real-time simulated crypto trading terminal — live candlestick chart, dynamic order book with flash animations, trade form modal, and auto-scrolling trade log.

#Dashboard#Trading#Chart#DataViz#RealTime#HighDensity

Evaluated 14 / 16

Pass Rate 8 / 13 (62%)

Avg Accessibility 40%

Score Range 0% – 100%

Tool Expert

Mini Vector Drawing Canvas Tool

Figma/Penpot-inspired vector editor — tool sidebar, pan/zoom canvas, layers panel, selectable/resizable shapes with handles, and SVG/PNG export.

#Tool#Canvas#Vector#Layers#SVG#DragAndDrop

Evaluated 15 / 16

Pass Rate 14 / 15 (93%)

Avg Accessibility 66%

Score Range 0% – 100%

Game Expert

Retro Arcade Game & Physics Engine

Polished neon/synthwave 2D arcade shooter — 60FPS requestAnimationFrame loop, canvas particles, Web Audio SFX, score/high-score/lives state, keyboard + touch controls.

#Game#Canvas#Physics#Web Audio#GameLoop#Particles

Evaluated 15 / 16

Pass Rate 15 / 15 (100%)

Avg Accessibility 95%

Score Range 75% – 100%

Audio Expert

Web Audio Synthesizer & Step Sequencer

16-step drum machine + synth pad built on pure Web Audio synthesis — 4 tracks, BPM/volume controls, playable pads, and a live visualizer.

#Web Audio#Audio#Sequencer#Synth#Canvas#Visualizer

Evaluated 15 / 16

Pass Rate 11 / 15 (73%)

Avg Accessibility 64%

Score Range 0% – 100%

Marketing Advanced

Scroll-Driven Storytelling Page

Cinematic scroll-driven product story for a smartwatch — parallax hero, GSAP/IntersectionObserver reveals, staggered feature animations, and pinned moments.

#Scroll#GSAP#Parallax#Motion#Storytelling

Evaluated 16 / 16

Pass Rate 15 / 16 (94%)

Avg Accessibility 70%

Score Range 0% – 100%

Marketing Intermediate

CodeSense Landing Page

Premium dark-glass landing page for an AI code-review tool — hero with animated headline, feature cards, testimonials, and a monthly/yearly pricing toggle.

#SaaS#Landing Page#Glassmorphism#CTA#PriceToggle

Evaluated 15 / 16

Pass Rate 14 / 15 (93%)

Avg Accessibility 62%

Score Range 0% – 100%

Dashboard Intermediate

Interactive Analytics Dashboard

Production-grade dark analytics dashboard — sidebar nav, KPI cards, Chart.js line + bar charts, and a date-range selector with a live activity feed.

#Dashboard#Chart.js#KPI#Responsive#DataViz

Evaluated 16 / 16

Pass Rate 15 / 16 (94%)

Avg Accessibility 56%

Score Range 0% – 93%

Coverage Matrix Suite

130 Runs 95% coverage

Visual evaluation matrix mapping prompt coverage and accessibility scores. Hover or focus a cell with multiple harnesses to compare them.

85%+ accessibility 75–84% accessibility <75% accessibility

Prompt Task	Category	DeepSeek V4 Flash DeepSeek 8/8	Gemini 3.6 Flash Google 8/8	GLM 5.2 Zhipu AI 8/8	GPT-5.6 Luna OpenAI 8/8	Grok 4.5 xAI 3/8	Hunyuan Hy3 Tencent 8/8	Kimi K3 Moonshot AI 8/8	Laguna S 2.1 Poolside 7/8	Ling 3.0 Flash Thinking Machines 8/8	Longcat 2.0 Meituan 8/8	Mimo V2.5 Xiaomi 8/8	Mimo V2.5 Pro Xiaomi 8/8	MiniMax M3 MiniMax 8/8	Muse Spark 1.2 Meta 8/8	Qwen 3.8 Max Alibaba 8/8	Step 3.7 Flash StepFun 8/8
Digital Garden Creative • Generative · 16/16	Interactive	96% ZCode	96% Antigravity CLI	95% ZCode	95% Codex	96% Grok CLI	92% ZCode	96% ZCode	92% ZCode	94% OpenCode	94% ZCode	96% ZCode	92% ZCode	96% ZCode	0% Command Code	98% ZCode	88% ZCode
High-Density Crypto Trading Terminal Dashboard • Trading · 14/16	Dashboard	0% ZCode	83% Antigravity CLI	100% ZCode	98% Codex	Not evaluated	88% ZCode	A11y N/A Command Code	Not evaluated	0% ZCode	3% OpenCode	0% ZCode	14% ZCode	0% ZCode	0% Command Code	90% ZCode	41% ZCode
Mini Vector Drawing Canvas Tool Tool • Canvas · 15/16	Tool	87% ZCode	90% Antigravity CLI	58% ZCode	73% Codex	Not evaluated	94% ZCode	66% ZCode	85% ZCode 2v	97% ZCode	43% ZCode	26% ZCode	22% ZCode	78% ZCode	0% Command Code	53% ZCode	96% ZCode
Retro Arcade Game & Physics Engine Game • Canvas · 15/16	Game	98% ZCode	83% Antigravity CLI	96% ZCode	75% Codex	Not evaluated	96% ZCode	100% ZCode	100% ZCode	98% ZCode	96% ZCode	100% ZCode	88% ZCode	100% ZCode	98% Command Code	100% ZCode	93% ZCode
Web Audio Synthesizer & Step Sequencer Web Audio • Audio · 15/16	Audio	98% ZCode	97% Antigravity CLI	88% ZCode	0% Codex	Not evaluated	68% ZCode	100% ZCode	100% ZCode	25% ZCode	85% ZCode	0% ZCode	0% ZCode	98% ZCode	95% Command Code	100% ZCode	0% ZCode
Scroll-Driven Storytelling Page Scroll • GSAP · 16/16	Marketing	100% 4 harnesses DeepSeek V4 Flash on Scroll-Driven Storytelling Page 100% Cline v1.0 · latest 55% Command Code v1.0 · latest 69% OpenCode v1.0 · latest 100% ZCode v1.0 · latest Launch in Harness Comparator	98% Antigravity CLI	55% ZCode	95% Codex	65% Grok CLI	100% ZCode	45% ZCode	98% ZCode	68% ZCode	50% ZCode	39% ZCode	95% ZCode	62% ZCode	0% Command Code	90% Command Code	58% ZCode
CodeSense Landing Page SaaS • Landing Page · 15/16	Marketing	100% OpenCode	0% Antigravity CLI	33% ZCode	100% 2 harnesses GPT-5.6 Luna on CodeSense Landing Page 100% Codex v1.0 · latest 63% ZCode v1.0 · latest Launch in Harness Comparator	Not evaluated	91% ZCode	95% ZCode	41% Command Code	35% OpenCode	88% ZCode	13% OpenCode	13% 2 harnesses Mimo V2.5 Pro on CodeSense Landing Page 13% OpenCode v1.0 · latest 25% ZCode v1.0 · latest Launch in Harness Comparator	80% Command Code	80% Command Code	93% ZCode	51% ZCode
Interactive Analytics Dashboard Dashboard • Chart.js · 16/16	Dashboard	39% OpenCode	83% Antigravity CLI	88% ZCode	80% ZCode	11% Grok CLI	90% ZCode	0% GitHub Copilot	5% 2 harnesses Laguna S 2.1 on Interactive Analytics Dashboard 5% Command Code v1.0 · latest 19% OpenCode v1.0 · latest Launch in Harness Comparator	19% OpenCode	54% ZCode	64% Cline	64% 2 harnesses Mimo V2.5 Pro on Interactive Analytics Dashboard 64% Cline v1.0 · latest 39% ZCode v1.0 · latest Launch in Harness Comparator	88% ZCode	93% Command Code	55% ZCode	44% Kilo