Contextual AI has released a new grounded language model (GLM) that reportedly achieves 88% factual accuracy on the FACTS benchmark, outperforming offerings from Google, Anthropic, and OpenAI. According to Michael Nuñez from VentureBeat, the startup was founded by pioneers of retrieval-augmented generation (RAG) technology. The new model specifically targets enterprise applications where factual precision is crucial. CEO Douwe Kiela emphasized that their approach differs from general-purpose models by prioritizing accuracy over creative flexibility. The company’s “RAG 2.0” system jointly optimizes all components rather than connecting off-the-shelf parts. Contextual AI has secured customers including HSBC, Qualcomm, and The Economist, and recently added support for multimodal content including charts and structured data from platforms like BigQuery and Snowflake.