Astra Pro nearly ties Claude Fable 5.1 on SimpleBench, a user reports

A user says GPT-6 Astra is on SimpleBench, reporting that Astra Pro scored 86.5% against Claude Fable 5.1’s 86.6%, with both beating the benchmark’s 83.7% human baseline. The post describes SimpleBench as testing everyday reasoning—from spatial logic to social situations and trick questions—and frames the results as progress on seemingly simple questions that have tripped up capable AI models.