Jerry Liu Benchmarks Fable 5.1 on ParseBench

Jerry Liu, co-founder and CEO of LlamaIndex, posted that Fable 5.1 arrived without document OCR benchmarks. He stated that the team therefore ran the model through ParseBench, their parsing benchmark covering real-world documents from finance, legal, insurance and other fields. The post notes strong OCR performance relative to recent frontier models. ParseBench focuses on tables, charts, content faithfulness, semantic formatting and visual grounding. The site parsebench.ai presents it as a tool to evaluate how well parsers work for AI agents, while LlamaIndex maintains a related cloud service at cloud.llamaindex.ai.