+1 (415) 347-6981Get started
Product & Engineering
Brand & Design
AI & Automation
AI Security
AI Labs
Ongoing
Fixed prices · You own the code · Proposal in 24h hello@startrise.io →
Startrise LLM Benchmark · 5 runs

Every build, every model

353 deliverables from 30 models across 12 briefs, exactly as the models returned them. Pick a brief to see the whole field side by side.

Explore the GitHub repo ↗
  1. Claude Fable 5.1 - Three.js Scroll Journey, the leading build
    Three.js Scroll Journey Frontend — extreme
    1. Claude Fable 5.1 91.4
    2. 02 Claude Opus 5 85.8
    3. 03 Kimi K3 84.3
    30 buildsBrowse all
  2. Claude Fable 5 - Custom WebGL Shader, the leading build
    Custom WebGL Shader Frontend — creative extreme
    1. Claude Fable 5 87.3
    2. 02 Qwen 3.8 Max 84.6
    3. 03 Claude Sonnet 5.5 84.6
    30 buildsBrowse all
  3. Qwen 3.8 Max - Brand Landing Page, the leading build
    Brand Landing Page Frontend — design
    1. Qwen 3.8 Max 87.3
    2. 02 Claude Opus 5.5 86.5
    3. 03 Claude Opus 5 85.8
    30 buildsBrowse all
  4. Claude Opus 5 - Playable 3D Game, the leading build
    Playable 3D Game 3D / game
    1. Claude Opus 5 88.0
    2. 02 Claude Sonnet 5.5 86.5
    3. 03 Claude Sonnet 5.5 (high) 85.0
    30 buildsBrowse all
  5. Claude Fable 5 - Open Creative Brief, the leading build
    Open Creative Brief Creative — open
    1. Claude Fable 5 85.7
    2. 02 GPT-5.6 Sol 83.8
    3. 03 Claude Opus 5 83.5
    29 buildsBrowse all
  6. Claude Opus 5.5 - Sell Yourself, the leading build
    Sell Yourself Persuasion & self-knowledge
    1. Claude Opus 5.5 83.8
    2. 02 Claude Opus 5.5 (high) 81.9
    3. 03 Kimi K3 80.2
    30 buildsBrowse all
  7. Claude Sonnet 5.5 (high) - Accessible Interface, the leading build
    Accessible Interface Interface engineering — accessibility
    1. Claude Sonnet 5.5 (high) 90.4
    2. 02 GPT-6 Astra 90.1
    3. 03 Claude Opus 5.5 (high) 87.1
    30 buildsBrowse all
  8. Claude Opus 5 - Brownfield Change Request, the leading build
    Brownfield Change Request Engineering discipline — maintenance
    1. Claude Opus 5 81.1
    2. 02 GPT-6 Astra 79.7
    3. 03 Kimi K3 79.6
    30 buildsBrowse all
  9. Muse Spark 1.3 - SVG Icon System, the leading build
    SVG Icon System Design & taste — iconography
    1. Muse Spark 1.3 83.9
    2. 02 Qwen 3.8 Max 81.6
    3. 03 Claude Fable 5 80.5
    27 buildsBrowse all
  10. GPT-6 Astra - Stateful Application, the leading build
    Stateful Application Interface engineering — state
    1. GPT-6 Astra 82.0
    2. 02 GPT-6.1 Sol 80.7
    3. 03 Claude Sonnet 5.5 (high) 79.9
    29 buildsBrowse all
  11. GPT-6.1 Sol - Zero JavaScript, the leading build
    Zero JavaScript Design & taste — CSS
    1. GPT-6.1 Sol 88.3
    2. 02 Claude Opus 5.5 (high) 86.4
    3. 03 Claude Opus 5 84.8
    28 buildsBrowse all
  12. Grok 4.6 - HTML Email, the leading build
    HTML Email Engineering discipline — email
    1. Grok 4.6 87.6
    2. 02 Claude Opus 5 84.3
    3. 03 Claude Opus 5.5 84.3
    30 buildsBrowse all

Covers are the all-time leader of each brief. Inside, every build is one card, sortable by score, human rating, panel or tokens, filterable by lab and run. Back to the leaderboard