OpenAI Thursday was released its smaller and more affordable version flagship large language model It powers ChatGPT. The new model, called the GPT-4o mini, will cost developers 60 percent less to build AI-powered apps and services than Open’s smallest model to date, the GPT-3.5 Turbo. But the big news here is for consumers. GPT-4o mini will replace GPT-3.5 Turbo for free ChatGPT users starting today – which means your basic ChatGPT experience will be greatly improved.
OpenAI claimed that the GPT-4o mini scored 82 percent on an industry benchmark called MMLU, which stands for Measurement of Massively Multitasking Language Understanding, and includes 16,000 multiple-choice questions across 57 academic subjects. This is slightly lower than the GPT-4o, which scored 88.7 percent, but higher than the GPT-3.5 Turbo, which scored only 70 percent. AI experts warned Against relying on such benchmarks to measure how intelligent AI systems are, however, so far they are the only way to measure the performance of large language models.
The smaller versions of the flagship models give developers more flexibility when it comes to building AI-powered apps. After all, not everyone needs or can afford access to the biggest, most powerful models for every use case. Google in May Gemini announced Flash 1.5, its lightweight model that the company says is optimized for speed and efficiency. Other AI companies, such as Anthropic, also have smaller versions of their full-scale models.
The GPT-4o mini can currently accept and generate text and images, but the model will eventually be able to process other types of content, such as audio and video. While GPT-3.5 will move away from Turbo ChatGPT, developers can still access it through OpenAI’s API to build apps and services with it – until it’s gone.
