OpenAI triples GPT-5.6 Sol's ARC score by preserving its memory
The 38.3% result used a custom public-set harness, while Claude Opus 5 remains first on ARC Prize's verified leaderboard.
By Ryan Merket · Published
Primary source: X - Tibo
Why it matters
OpenAI's result shows that memory and context plumbing can triple an agent's benchmark score, complicating leaderboards and making orchestration a core product advantage.
OpenAI core products lead Thibault "Tibo" Sottiaux (@thsottiaux) said GPT-5.6 Sol reached a state-of-the-art score on ARC-AGI-3 after OpenAI changed how the model's reasoning and context were carried between actions. The claim rests on a custom run of the benchmark's public tasks rather than ARC Prize's official verified leaderboard, where Anthropic's Claude Opus 5 remained the leader as of July 29th.…
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.