GPT-4V(ision) System Card

Abstract

This system card describes GPT-4 with vision (GPT-4V), which enables users to instruct GPT-4 to analyze image inputs. We describe the safety evaluations, mitigations, and deployment preparation for the multimodal capabilities of GPT-4V including visual question answering, image description, spatial reasoning, and document understanding.

Key Findings

• 1Extended GPT-4 with robust image understanding capabilities
• 2Demonstrated strong visual question answering and spatial reasoning
• 3Handled diverse image types: photos, diagrams, charts, screenshots, documents
• 4Included extensive safety evaluations for multimodal risks
• 5Enabled new applications in accessibility, education, and analysis

Impact & Significance