Artificial Analysis benchmarked AI research tools Astra and Fable at maximum effort, finding task-dependent performance differences but no clear winner for reliable live research or citation accuracy. Astra outperformed on a PDF comprehension test, while fallback-enabled Fable scored higher on a quiz that penalized incorrect answers. Despite these individual strengths, neither tool demonstrated consistent reliability in conducting live research or producing accurate citations.