ChatGPT-4

The advanced multimodal LLM from OpenAI: GPT-4 processes text and image inputs, offering human-level performance on academic and professional benchmarks (e.g., top 10% on the Uniform Bar Exam).

GPT-4 is OpenAI’s large multimodal model (LMM), representing a significant capability leap from GPT-3.5. This technology accepts both text and image inputs, enabling complex reasoning tasks like explaining a graph or debugging code from a screenshot. The model demonstrates human-level proficiency, scoring in the 90th percentile on the Uniform Bar Exam, a massive improvement over GPT-3.5’s bottom 10% performance. Furthermore, it manages a drastically increased context length—up to 32,768 tokens (approximately 25,000 words)—and is 82% less likely to generate disallowed content, ensuring greater utility and safety for high-stakes applications.

https://openai.com/gpt-4