Main Purpose
GPT-4 is a multimodal large language model created by OpenAI.
It is the fourth model in the GPT series and is designed to generate, edit, and iterate on creative and technical writing tasks.
Key Features
- GPT-4 is more reliable, creative, and able to handle more nuanced instructions compared to its predecessor, GPT-3.5.
- It has two versions with context windows of 8,192 and 32,768 tokens, a significant improvement over previous models.
- GPT-4 is a multimodal model, capable of taking images as well as text as input.
- It can describe humor in unusual images, summarize text from screenshots, and answer exam questions that contain diagrams.
- GPT-4 can interact with external interfaces, perform web searches, use APIs, generate images, and access and summarize webpages.
- It can be given a system message to specify its tone of voice and task, allowing for more control over its responses.
- GPT-4 has demonstrated aptitude on standardized tests, scoring well on exams such as the SAT, LSAT, and Uniform Bar Exam.
- It has been found useful for assisting in coding tasks, finding errors in code, and suggesting optimizations.
- GPT-4 has been tested on medical problems and has shown promise in the medical field.
Use Case
- Creative and technical writing tasks
- Generating songs and screenplays
- Learning a user's writing style
- Assisting in coding tasks
- Medical applications
