OpenAI, the company behind the popular ChatGPT, introduced a new GPT-4o mini model on Thursday. It’s the latest small AI model, which the company says is cheaper and faster than other top-of-the-line OpenAI models. The new AI model is available now to developers and users via the ChatGPT web or mobile app (enterprise users will get access next week).
According to the company in post on the website the GPT-4o mini outperforms the leading small AI models in tasks involving text or images, while costing significantly less and being more cost-effective (more than 60% cheaper than the GPT-3.5 Turbo) compared to larger models such as the GPT-4o or Claude 3.5 Sonnet from Anthropic. This will make it more usable, for example, for developers who use small AI models to solve a large volume of simple tasks.
The GPT-4o mini will replace the existing GPT-3.5 Turbo as the smallest model offered by OpenAI. From today it is available on the Free, Plus and Team tariffs. Enterprise users will get access to it as well as enterprise users next week. GPT-3.5 will no longer be available to normal users, it will still be available through the developer API for some time.
The new lightweight model can currently work with text and images, but according to the company, it should soon switch to fully multimodal inputs and outputs (text, video and audio). The GPT-4o mini scored 82% on the MMLU (Measuring Massive Multitask Language Understanding) benchmark for measuring logical reasoning.
Comparison of AI models
|
A comparison of small AI models from artificial analysis to their price ratio
|
The price for developers who want to build their applications through the OpenAI API is set at 15 cents per million input tokens and 60 cents per million output tokens. The GPT-4o mini thus has a total of 128,000 tokens, which roughly corresponds to the length of 2,500 pages in a standard book.
Along with the announcement of the new GPT model, OpenAI announced according to the magazine TechCrunch new tools for enterprise customers such as Enterprise Compliance API. This tool helps businesses in highly regulated industries meet logging and auditing requirements. It enables administrators to monitor and control data interactions within ChatGPT Enterprise, increasing security and control over AI use in enterprises.
Sources: OpenAI (1, 2), TechCrunch, Artificial Analysis
Source: www.cnews.cz
