Modeles traites
7
Complets
6
Non complets
1
Attendus
33
Resultats
| Modele | Provider | Model ID | Status | Duree | Session | Note |
|---|---|---|---|---|---|---|
| 01. Owl Alpha (stealth) [28/04/2026] | OpenRouter | OpenRouter-Free-MAJ/openrouter/owl-alpha |
complete | 00:27:09 | ses_20af31a60ffeilkWx5biis1oBc | {"type": "error", "timestamp": 1777957436997, "sessionID": "ses_20af31a60ffeilkWx5biis1oBc", "error": {"name": "UnknownError", "data": {"message": "\"Provider returned error\""}}} |
| 03. Poolside: Laguna XS.2 [28/04/2026] | Poolside | OpenRouter-Free-MAJ/poolside/laguna-xs.2:free |
complete | 00:01:06 | ses_2097a43b9ffeuMUtOP5W231JNp | |
| 04. Poolside: Laguna M.1 [28/04/2026] | Poolside | OpenRouter-Free-MAJ/poolside/laguna-m.1:free |
complete | 00:02:49 | ses_209793edeffe1BHsJBkXevOkOr | |
| 05. inclusionAI: Ling-2.6-1T [23/04/2026] | inclusionAI | OpenRouter-Free-MAJ/inclusionai/ling-2.6-1t:free |
parse error | 00:03:52 | ses_20976a80effe5pAB4fwsHawBqP | node --check failed |
| 06. Tencent: Hy3 preview [22/04/2026] | Tencent | OpenRouter-Free-MAJ/tencent/hy3-preview:free |
complete | 00:08:26 | ses_209604547ffe6RpQN2KKi03NTN | |
| 17. NVIDIA: Nemotron 3 Nano 30B A3B [14/12/2025] | NVIDIA | OpenRouter-Free-MAJ/nvidia/nemotron-3-nano-30b-a3b:free |
complete | 00:03:25 | ses_2094ef636ffeUJiYdwPYmif72I | |
| 21. OpenAI: gpt-oss-120b [05/08/2025] | OpenAI | OpenRouter-Free-MAJ/openai/gpt-oss-120b:free |
complete | 00:03:14 | ses_2094a5b37ffeHNk5r5JBfY5v75 |
Prompt de reprise
Wrapper
Suis strictement le fichier joint. Reponds en francais. Tu travailles dans le dossier courant et tu dois generer toi-meme les fichiers reels. Contraintes non negociables : - Cree directement `index.html`, `styles.css` et `app.js` dans le dossier courant. - Ne mets pas leur contenu dans ta reponse finale. - N'ajoute pas d'autres fichiers si tu peux l'eviter. - Quand les 3 fichiers existent, termine par un court message de confirmation en une ou deux lignes maximum.
Benchmark source
# Full Prompt: Generate a production-level "Dinosaurs" web page (design + verifiable calculations + docs)
## ๐ฏ Role & Objective
You are a senior **agentic AI** (full-stack + data + UX) tasked with delivering a **static self-contained web page** about **dinosaurs**, demonstrating:
1. **Premium design** (elegant, modern, accessible).
2. **Non-trivial scientific calculations, traceable and verifiable** (with sources and tests).
3. **Clear documentation** (technical, functional, and scientific).
The result must be **directly usable** (opening `index.html` works without backend). No external build is required, but if you suggest a mini-build (e.g. Vite/Parcel), also provide a "dist" version ready to use.
---
## โ ๏ธ Strict Constraints (3-file format)
* **Only** `index.html`, `styles.css`, `app.js`.
* **No external dependencies** (no CDN/NPM, no network calls). System fonts only.
* **Offline-first**: opens locally.
* **Data (CSV)**, **constants (JSON)** and **docs (Markdown)** must be **embedded inside `app.js`** as template strings. A minimal Markdown parser converts content to HTML.
* Images/illustrations must be **inline SVGs** (in `index.html`) or dynamically generated SVGs via JS.
---
## ๐งฑ Expected Deliverables
* Exactly **3 files** (simplified MVC):
1. `index.html` โ **View**: semantic structure, UI components, panels (Explorer, Methods & Calculations, Docs, Tests), Export buttons.
2. `styles.css` โ **Style**: premium design (light/dark), grid, tokens, components (Cards, Chips, Tabs, Modal, Tooltip), accessibility (focus + contrast).
3. `app.js` โ **Controller + Model**:
* **Model**: dinosaur dataset (CSV **embedded as template string**), scientific constants (JSON **embedded**), minimal CSV parser, schema validation.
* **Calculations**: `massFromFemur`, `bmr`, `speedFromFroude`, `toTimelinePosition`, `explain()`.
* **Controller**: global state, filters, sorting, search, lightweight hash routing, test execution, JSON/CSV/MD export.
* **Integrated tests**: assertions, visual reporting, tolerances, "Run Tests" button.
* **No other resources**: no `data/`, `docs/`, or `tests/` folders. All documentation is **embedded as Markdown strings in `app.js`** and rendered in `index.html`.
---
## ๐งญ Page Content & Structure
1. **Hero section** (title, baseline, CTA to calculations). Subtle visual background (gradient, SVG noise, mesh, or low-poly WebGL).
2. **Dinosaur Explorer**: browsable/filterable gallery (geological period, diet, clade, estimated mass, length). Real-time search.
3. **Timeline** of eras (Triassic/Jurassic/Cretaceous) with species positions (ยฑ uncertainty). Smooth animation, keyboard controls.
4. **Species Sheets** (modal or anchor page): name, clade, period, measured data (length, femur circumference if available), **derived calculations** (estimated body mass, theoretical BMR, plausible speed, growth), sources.
5. **"Methods & Calculations" Section**: formulas, assumptions, limits. Highlighted "Key Insight" boxes.
6. **Charts**: distributions (histograms, KDE approximation), allometric relations (log-log), ternary plots if relevant. Interactive (hover, zoom, tooltips, export PNG/SVG).
7. **Species Comparator** (multi-selection).
8. **FAQ** and **Glossary**.
9. **Footers**: credits, license, links to technical docs, "View/Run Tests" button.
---
## ๐จ Design & UX Requirements
* **Premium editorial style**: 12-column grid, 16โ24px border radius, soft shadows, generous spacing, clear typography hierarchy. **Light/dark mode**.
* **Design system** in `DESIGN_SYSTEM.md`: colors (HSL), typography, scales, radii, shadows, components (Card, Tag, Chip, Badge, Tooltip, Modal, Tabs, Toast).
* **Accessibility**: WCAG 2.2 AA compliant, contrast tested, visible focus, ARIA roles, keyboard navigation, `prefers-reduced-motion` respected.
* Subtle micro-interactions at 60fps.
---
## ๐ Required Data Format
* **Embedded CSV**: `<script type="text/csv" id="dinosaurs_csv">` with โฅ30 species.
* **Embedded JSON constants**: `<script type="application/json" id="constants_json">` with references (title, author, year, DOI/URL).
* Missing values allowed but **uncertainty must be displayed** visually.
* Provide **UI controls** to adjust constants live (sliders/inputs).
---
## ๐งฎ Required Calculations (Verifiable)
Provide **pure JS** implementations + **unit tests**.
1. **Body mass from femur circumference** (allometry, log-log):
* General form: `log10(M_kg) = a + b * log10(C_mm)` โ `M_kg = 10^(a + b*log10(C_mm))`
* Read `a`, `b` from constants.
* Must include **reference** + sample check panel.
2. **Basal Metabolic Rate (BMR)**:
* Compare **endotherm vs ectotherm models** via slider.
3. **Plausible speed** (Froude number):
* `Fr = v^2 / (g*l)` โ `v = sqrt(Fr * g * l)`
* Leg length estimation: `0.4 * body length`.
4. **Timeline conversion**:
* Convert `age_mya_start/end` into interactive timeline position.
5. **Uncertainty propagation** *(optional but great)*.
Each model must:
* Provide `explain(modelName, inputs)` โ returns steps + sources.
* Include **unit tests**.
---
## ๐ Verifiability & Transparency
* Button **"Verify Calculations"** runs a test suite with clear output.
* Button **"Export Proofs"** outputs a JSON with constants + inputs + result.
* Each calculation must show a **step-by-step detail view**.
---
## ๐งฐ Stack & Tech Constraints
* **No heavy dependencies**. For charts, use SVG or embed a tiny local library.
* No frameworks required (optional Preact/HTM/Web Components OK).
* Optional TypeScript locally compiled into JS (include compiled version).
* **No network requests** at runtime.
---
## ๐งช Quality & Testing (Integrated)
* Inline tests for `massFromFemur`, `bmr`, `speedFromFroude`, `toTimelinePosition`.
* CSV schema validation.
* Mini test runner UI with JSON export.
* Reasonable size limit.
---
## โฟ Accessibility & i18n
* ARIA roles, labels, tab order.
* Automatic checks (e.g. required button labels).
* Content **in French**, but i18n-ready.
---
## ๐ Performance & SEO
* Lighthouse โฅ95 target.
* Lazy loading, optimized SVG, local fonts with `font-display: swap`.
* Meta tags + Open Graph + JSON-LD.
---
## ๐งพ Embedded Documentation (in `app.js`)
* Markdown strings: `README`, `ARCHITECTURE`, `MATH`, `DESIGN_SYSTEM`, `ACCESSIBILITY`.
* Download as `.md`.
---
## ๐ Licenses & Credits
Prefer CC-BY, CC0, MIT.
---
## โ
Checklist
* Offline functional
* Premium responsive design (dark/light)
* Dataset โฅ30 species
* Documented calculations
* Tests integrated
* Accessibility + SEO
* Full documentation
---
## ๐ Inline Templates/Examples
(CSV + JSON examples preserved here as in original prompt)
---
## ๐ Required UI Interactions
* Multi-filter chips
* Sorting
* Export CSV
* Toggle light/dark mode
* Test runner
---
## ๐ Bonus (optional)
* Hypothesis comparison (endo vs ecto)
* Uncertainty visual bands
* PWA
* SVG sprite generation
---
## ๐ฃ Final Instructions
* Zero network usage.
* Show calculation traceability.
* Deliver full folder ready to open.
**Expected Output**:
1. Directory structure summary
2. Full content of `index.html`, `styles.css`, `app.js`
3. Inline data/docs/test content
4. Simple instructions to run
Le prompt benchmark original est conserve dans prompt.txt. Les journaux bruts de cette execution vivent dans _runs/260505/.