ARC-AGI-3 Difficulty Distribution and Model Scores
GregKamradt · x · 2026-07-10
The post shares observations on the difficulty of ARC-AGI-3 public games sorted by GPT-5.6 Sol (Max) scores: the easiest levels include ar25 (38%) and ft09 (87%), while the hardest levels are g50t, sk48, and su15, all scoring 0%.
Essentially, this information uses specific level scores to illustrate the model's performance variance on ARC-AGI-3, categorizing it as model evaluation/leaderboard content.
Related event: GPT-5.6 Sets New SOTA on ARC-AGI-3 and Exceeds 30% on GDP.pdf(16 posts)→
More from Models
- Claim says Kimi was distilled from Fable, sparking a model-attribution jab — cephaloform · 2026-07-22
- Gary Marcus says LLM math skills are like knowing only a car’s engine size — GaryMarcus · 2026-07-22
- OpenAI’s Codex + GPT-5.6 Sol hits 99% recall in Project APE verification tests — soumitrashukla9 · 2026-07-22
- OpenAI-linked paper says capability RL can make models more reward-seeking — MariusHobbhahn · 2026-07-22
- Macaron V1 adds LoRA RL on GLM 5.2 and claims SOTA benchmark gains — Xianbao_QIAN · 2026-07-22
- OpenAI rolls out voice in GPT-Live, but the UI obscures search and reasoning — Graham_dePenros · 2026-07-22