Open leaderboard

Run of Nov 2

Accuracy on 1,400 multi-step problems

  1. Halcyon Pro, 84.1%
  2. Gannet XL, 81.7%
  3. Wren 3 (you), 79.4%
  4. Osprey 70B, 76.2%
  5. Marlin 2, 72.8%
  6. Petrel Mini, 61.5%