Anthropic's Opus 5.5 model scored higher on the Artificial Analysis Imagination index than Fable 5.1 and GPT-6 Astra