OpenAI triples GPT-5.6 Sol's ARC-AGI-3 puzzle benchmark scores from 13.3% to 38.3% by enabling two API settings—retained reasoning and compaction

OpenAI triples GPT-5.6 Sol's ARC-AGI-3 puzzle benchmark scores from 13.3% to 38.3% by enabling two API settings—retained reasoning and compaction—while cutting output tokens by 6x.

How enabling two settings tripled our scores on the ARC-AGI-3 benchmark

A sped-up video of GPT‑5.6 Sol attempting to solve puzzles in the ARC-AGI-3 benchmark, with the official harness (left) and our Responses API harness (right), which retains reasoning and enables compaction. On the leaderboard for this game⁠(opens in a new window), no frontier model solves any level beyond the first. With our harness, GPT‑5.6 Sol solves all six.GPT‑5.6 Sol has solved longstanding open problems in mathematics like the cycle double cover conjecture⁠(opens in a new window) and beate...