In the official announcement, OpenAI, has introduced GPT-4o Mini, a new iteration in the GPT-4 series designed to deliver high-quality AI performance while being significantly more cost-effective. This latest model aims to make advanced AI capabilities more accessible and affordable for a wider range of applications and users.

Key Features of GPT-4o Mini:

1. Cost Efficiency: GPT-4o Mini is engineered to reduce operational costs without compromising on the quality of output. This makes it an ideal solution for businesses and developers seeking budget-friendly AI solutions. We can see that GPT-4o mini is more than 60% cheaper than GPT-3.5 Turbo, priced at 15¢ per 1M input tokens and 60¢ per 1M output tokens (roughly the equivalent of 2500 pages in a standard book).
2. Compact Size, High Performance: Despite its smaller footprint, GPT-4o Mini maintains robust performance metrics, ensuring that users can still leverage the powerfol capabilities of the GPT-4 architecture. It is stated in the official announcement that GPT-4o mini outperforms GPT-2.5 Turbo in textual intelligence-scoring 82% on MMLU compared to 69.8% and multimodal reasoning
3. Versatility: The model is designed to cater to various applications, from customer service bots and content generation to complex data analysis, proving its adaptability across different sectors.
4. Accessibility: By lowering the cost barrier, OpenAI aims to democratize AI technology, making it more accessible to startups, small businesses, and educational institutions.
5. Environmental Impact: The efficient design of GPT-4o Mini also translates to lower energy consumption, aligning with sustainable practices and reducing the carbon footprint of AI operations.
6. Modalities: GPT-4o mini currently supports text and vision capabilities, and the team stated that it is planned to add support for audio and video inputs and outputs in the future.
7. Language: GPT-4o mini has improved multilingual understanding over GPT-3.5 Turbo across a wide range of non-English languages. Take a second and test your own mother tongue it may surprise you positively!

Other than that jeffsharris (OpenAI Staff) stated that:

Like GPT-4o, GPT-4o mini has a 128k context window and a knowledge cut-off date of October 2023. It also supports up to 16,384 max_tokens 82.

We plan to launch fine-tuning for GPT-4o mini in the coming days. You can learn more about GPT-4o mini in our announcement blog 659 and API documentation 532, or by testing the model in Playground 90. Excited to hear what you think!

This release marks a significant step in OpenAI's mission to provide powerfol, efficient, and accessible AI tools, empowering more users to harness the potential of artificial intelligence in their respective fields.

For those of you using OpenAI models through Azure, worry not, in this announcement that was published on 18th of July 2024, Microsoft has announced the availability of OpenAI's latest and fastest model, GPT-4o Mini, on Azure AI. This integration brings the advanced capabilities of GPT-4o Mini to Azure users, offering enhanced performance and cost efficiency. This collaboration between Microsoft and OpenAI underscores a commitment to making advanced AI more accessible and affordable, empowering businesses to innovate and optimize their operations with the latest AI tools.

Update 27/07/24 : In order to fine-tune the model and use it for your own applications, take a look at the official fine-tuning instructions, here