Open BurnBar
GitHub Download
Model board · the archive

Every board verdict, frozen, cited, and inspectable.

BurnBar refreshes its model-landscape snapshot once every twenty-four hours, then a board of language models runs research and analysis tasks over it. Each refresh produces a dated rundown — what BurnBar would have recommended that day, the sources it cited, the freshness of every signal, and the limitations of every claim. Nothing is overwritten. Future humans can verify what was said.

  • 12 dated rundowns
  • newest · 2026-06-12
  • oldest · 2026-05-13

Benchmark signals are advisory. Each rundown shows what the daily model board recommended, not what the router actually ran — provider-family mode, user pinning, account auth, quota state, safety policy, availability, and exact-model failover rules all override these recommendations at runtime. Use this archive to inspect why a given recommendation was made, not as proof a given route was executed.

2026-05

7 rundown s

Sources we cite

Each daily rundown attributes its claims back to a documented public source. Citations carry logos so operators can audit provenance at a glance. We never fabricate benchmark numbers, scrape authenticated dashboards, or persist API keys inside snapshots.