A new benchmark called Agents' Last Exam has been launched to measure artificial intelligence's ability to execute professional workflows. This benchmark has been used to compare the performance of different AI models, with GPT-5.5 achieving the top spot. GPT-5.5 beat Claude Fable 5, a highly anticipated model, with a pass rate of 24.0% compared to 22.0%.

AI summaries can be wrong sometimes—always verify important details using the source article.