OpenAI has unveiled a smaller, more cost-effective version of its most powerful model, GPT-4o, which could enable even more AI-powered apps.
The new model is GPT-4o MiniAccording to OpenAI, GPT-4o mini is smarter than GPT-3.5 Turbo and over 60% cheaper, making it more affordable for developers building applications with OpenAI's models. GPT-4o mini costs 15 cents per million input tokens (the units of information used by AI models, such as words or phrases) and 60 cents per million output tokens, which OpenAI says is equivalent to 2,500 book pages.
“We expect GPT-4o mini will make intelligence much more affordable and greatly expand the range of applications built with AI,” OpenAI said in a statement.
According to OpenAI, GPT-4o mini scored 82% on the Massive Multitask Language Understanding (MMLU) benchmark, which measures a language model's ability in 57 subjects including mathematics and law, outperforming other mini models and GPT-3.5 Turbo. The mini model replaces GPT-3.5 Turbo in ChatGPT, and Free, Plus and Team users can now access the new model.
At the time of release, GPT-4o mini supports text and vision in its API (application programming interface), allowing developers to use OpenAI's technology to build their own AI models. In the future, OpenAI says GPT-4o will support text, images, video and audio. The mini model has knowledge up to October 2023.
“We envision a future where our models are seamlessly integrated into any app or website,” OpenAI said. “GPT-4o mini paves the way for developers to build and scale powerful AI applications more efficiently and affordably.”
Open AI Introducing ChatGPT-4oThe robot, due to be unveiled in May, will be multimodal, meaning it can see, hear and converse in real time.
