OpenAI Vision

Source: https://developers.openai.com/api/reference/python, https://github.com/openai/openai-python/blob/main/README.md?plain=1, https://api-docs.deepseek.com/

Purpose

Send images alongside text to vision-capable OpenAI models (gpt-4o, gpt-4o-mini, gpt-4-turbo) and OpenAI-compatible providers (DeepSeek, Qwen-VL, Zhipu GLM-4V). Images are passed as image_url content blocks within the messages array — either as HTTP URLs or base64-encoded data URIs.

When to Use

• Analyzing photos, screenshots, documents, charts, diagrams
• OCR — extracting text from images
• Visual Q&A and image classification
• Comparing multiple images in one request
• Combining vision with tool calling (classify → route)
• Cross-provider vision: DeepSeek-VL, Qwen-VL, GLM-4V via OpenAI-compatible endpoints

Basic Usage