GPT-5.6 Sol, the flagship model, achieves high scores on reasoning benchmarks like ARC-AGI-3, demonstrating enhanced planning and problem-solving in unfamiliar environments.
Open the full topic