With seven AI models and one assignment, authors John Tredennick and Lydiia Zhigmitova explore validation results around the capabilities of 7 AI models in respect to their accuracy, speed and cost. Anthropic's, Google's and OpenAI frontier models out in September 2026 were tested, including Astra and Fable for citations overall, citations at the paragraph level, fact accuracy and fidelity with surprising results.