Write a comprehensive wiki page for the new Model Benchmark System at /benchmarks.

Cover:

• Purpose: comparing models across categories (code-gen, code-review, reasoning, chat, summary, image-gen)
• Architecture: benchmark-types.ts, benchmark-storage.ts, benchmark-runner.ts, benchmark-seeds.ts
• How to use: New Run tab → pick category + prompt + models → Run → rate results → view Rankings
• Ranking formula: quality×0.5 + cost×0.2 + latency×0.15 + consistency×0.15
• Available models (all 22 across 9 providers)
• Free tier models highlighted
• Image generation comparison
• See: app/benchmarks/page.tsx, components/BenchmarkClient.tsx, lib/benchmark-*.ts, app/api/benchmarks/route.ts