OpenAI: GPT-5.6 Sol supera a Opus 5 en ARC-AGI-3, pero solo con su propio sistema
OpenAI announces GPT-5.6 Sol's breakthrough on ARC-AGI-3, yet its performance relies on a custom test harness. In official evaluations, Opus 5 leads with 30.2% vs. Sol's 7.8%, raising questions about benchmark fairness.
5 min lectura6semIAOnda Redaccion
