Verified Run by us through Codex, locked to the cube tools. Finished 26 Oct 2026 19:25 UTC.
GPT-6.1 Terra scores 12.6 at low, giving up on the interactive 2×2 ladder at level 3 with all 3 lives left. It falls well short of its best max result, 62.0, which puts it 3rd of 18 models on the Index. The low run is not the effort it is ranked on.
The 2×2 round ended after 16 s, with 2 cubes solved and no lives spent. On the interactive 3×3 ladder, it solved through level 3 before giving up on level 4, again with 3 lives left. The one-shot round reached the same point but included a failed level 4 answer before it gave up on a fresh cube with 2 lives remaining. Across the suite, it solved 8 cubes; none of the ladders ended through exhausted lives, and the Gauntlet stayed locked.
We timed the low run at 61 s, the fastest of its efforts. The medium run matched its score in 94 s, while the best max run took about 1 h 15 min. API-equivalent costs are unavailable for these runs.
At low, it solved through level 2, then gave up on level 3 without making a move, with all 3 lives left. Level 1 took 2 s and 1 move; level 2 accounted for 11 s of the 16 s round, using 6 moves, 4 more than the scramble. It stopped 3 s into the next cube.
At low, it solved through level 3 of 20, then gave up on level 4 with all 3 lives left. The first three cubes took 13 s combined, each solved in 1 move more than its scramble. On level 4, it played B B2 and stopped after 7 s, having used only 2 of the allowed 48 moves.
At low, it solved through level 3 of 20, then gave up on level 4 with 2 lives left. The first 3 answers exactly reversed their scrambles; level 3 took 12 s, half the round’s 24 s. Its first level 4 answer failed in 4 moves, and it gave up on the replacement cube without submitting any moves.