7 Things You Need to Know About OpenAI's GPT-4

More precise than GPT-3.5 - More precise than GPT3.5 (ChatGPT) with fewer flaws and hallucinations thanks to larger training data sets.

Visual Input Support Can describe images in great detail, like the Mona Lisa painting, but not yet available to the public.

GPT-4 Limitations Still struggles with recent events due to training data sets ending in late 2021.

Better Advanced Reasoning Capabilities Achieved greater accuracy than previous models by passing impressive exams such as the SAT and Bar exam. In a simulated Bar exam, it scored in the top 10% of test takers, while GPT3.5 scored in the bottom 10%.

Multilingual Support -Multilingual with the ability to translate and converse in about 24 languages, thanks to Microsoft Azure capabilities.

GPT-4 safer than GPT-3.5 -GPT-4 is safer than previous models, with a lower likelihood of responding to requests for disallowed content and a higher likelihood of producing factual responses, according to OpenAI's internal evaluations.