Setup required
Create secrets.json in the same folder as this file:
{
  "aa_api_key": "your-aa-key-here",
  "api_key":    "sk-or-v1-..."
}
Then restart the stack (docker compose restart model-analyser) and reload the page — secrets.json is mounted read-only, so it must exist on the host before the container starts. Keys are read server-side and never stored in the browser.

aa_api_key — required. Get a free key at artificialanalysis.ai → Insights Platform → API Keys.
api_key — optional OpenRouter key. Pricing data is public; the OR key is only needed for config download.
Model Analyser
Loading…
⚠ Artificial Analysis data unavailable — Coding Index and Value Ratio columns are empty. Check that the server container is up (docker compose ps) and your aa_api_key is set in secrets.json.
RouterForge
Benchmarks: Artificial Analysis (Coding · Intelligence · Agentic) · Pricing: openrouter.ai (live) ·
⚙ Tier Price Limits
Suggested: $0.50
Suggested: $1.50
Suggested: $3.00
1 = single model (no routing) · 10 = maximum spread
Optimise for
Model Pricing vs Quality Analysis
X-axis: output $/M tokens (log scale). Y-axis: Coding Index (Artificial Analysis). Top-left = best value. ◉ = your stack. Value ratio coding index ÷ $/M shown in tooltip.
Model Value Matrix — Comprehensive Comparison
✎ Click any row to add or remove from your stack. Changes are saved automatically.
# Model Tier Coding
Index (AA)
Intelligence
Index (AA)
Agentic
Index (AA)
Speed
tok/s (AA)
Value Ratio
coding index ÷ $/M
Output $/M
live OpenRouter
Input
modalities
Tools
calling · SO · JSON
Cache
prompt caching
Pin a model to your stack by ID (it's already in the table — this adds it to ★ your stack tracking):
Amber rows = your stack · models without AA benchmark data are hidden by default — use the No AA data toggle to show them ✓ Saved
Best-value S/M/L/P stack — Coding index ÷ price

Benchmarks: Artificial Analysis — Coding, Intelligence & Agentic indexes · source: AA API (primary), OpenRouter bundled data (fallback) · Pricing: openrouter.ai (live) · Attribution required per AA API terms.