OpenAI GPT-4

OpenAI's premier large multimodal model: accepts text and image inputs, delivers human-level performance (e.g., top 10% on the simulated bar exam), and processes up to 32,000 tokens.

GPT-4 is OpenAI's most advanced system: a large multimodal model that processes both text and images, emitting text outputs. This model demonstrates near human-level proficiency, notably scoring in the top 10% on the simulated Uniform Bar Exam (Source: 4). Its key operational advantage is the expanded context window: 32,000 tokens, enabling it to manage and generate approximately 25,000 words for complex, long-form applications (Source: 1). This represents a major scaling milestone in deep learning and general-purpose AI reasoning.

https://openai.com/research/gpt-4