Build a landing page from extracted design tokens
Exact prompt
Using ONLY the design tokens in tokens.json, build a landing page for "Ledgerly", an app that turns receipts into tax-ready expense reports. Output a single self-contained file index.html (inline CSS, no JS frameworks, Google Fonts link allowed). Required sections, each with the given id: hero (headline + CTA button), features (3 feature cards), pricing (2 plans), faq (3 questions), footer. Define every token as a CSS custom property in :root and use them. Do it in one pass.
Task prompt given to the agent (headless, offline, synthetic fixture inputs): Using ONLY the design tokens in tokens.json, build a landing page for "Ledgerly", an app that turns receipts into tax-ready expense reports. Output a single self-contained file index.html (inline CSS, no JS frameworks, Google Fonts link allowed). Required sections, each with the given id: hero (headline + CTA button), features (3 feature cards), pricing (2 plans), faq (3 questions), footer. Define every token as a CSS custom property in :root and use them. Do it in one pass. Inspired by https://x.com/micahtid/status/2082297045818175717 (unverified; not checked with the author). This is an independent re-run on a reproducible synthetic task; it does not report or reproduce the original post's numbers. Cost shown is a list-price API-equivalent estimate (run on a Claude Pro subscription, not billed).
Success criteria: All automated rubric checks pass: section id=hero present; section id=features present; section id=pricing present; section id=faq present; section id=footer present; all 6 color tokens used; both fonts used; radius/spacing token values used; tokens defined as :root custom properties; no off-token hex colors; no JS framework / renders without console errors.
Best model for this task so far: Qwen2.5 7B (local) (thinking off) (success, $0 (local), 50s). Best AI model for Software & DevOps →
All runs (10)
70% pass rate across 10 eligible runs: 7 pass, 2 partial, 1 fail. Partial and failed runs are listed below with everything else.ⓘ
| Model | Cost | Time | Step-insⓘ | Autonomy | Outcome | Trustⓘ | Date | Compare |
|---|---|---|---|---|---|---|---|---|
| MLlama 3.2 3B (local)thinking off | $0 (local) | 13s | 0 | Fully autonomous | fail | Trust 80% | 2026-10-20 | |
| QQwen2.5 7B (local)thinking off | $0 (local) | 50s | 0 | Fully autonomous | pass | Trust 80% | 2026-10-20 | |
| ZGLM 5.3 (Venice)thinking on | $0.06 | 1m | 0 | Fully autonomous | pass | Trust 80% | 2026-10-17 | |
| AClaude Haiku 5.5default · thinking on | $0.0061 | 26s | 0 | Fully autonomous | pass | Trust 80% | 2026-10-16 | |
| XGrok 4.3 (Venice) | $0.01 | 26s | 0 | Fully autonomous | partial | Trust 80% | 2026-10-15 | |
| QQwen3 Coder 480B Turbo (Venice) | $0.0057 | 34s | 0 | Fully autonomous | partial | Trust 80% | 2026-10-15 | |
| DDeepSeek V4.1 Flash (Venice) | $0.01 | 39s | 0 | Fully autonomous | pass | Trust 80% | 2026-10-14 | |
| GGemma 4 31B Instruct (Venice) | $0.0017 | 1m | 0 | Fully autonomous | pass | Trust 80% | 2026-10-14 | |
| AClaude Opus 5.5 | $0.30 | 1m | 0 | Fully autonomous | pass | Trust 80% | 2026-10-14 | |
| AClaude Sonnet 5.5 | $0.09 | 26s | 0 | Fully autonomous | pass | Trust 80% | 2026-10-14 |
Embed this result
A live badge for this use case. It links back here and updates as new runs are published.
Get new runs and ranking changes weekly
Which models got better or cheaper on real tasks like this one.