GPT-4o Mini Is Now Available on ChatGPT for Free. OpenAI Is Using Muscle Against Google and Anthropic With This Small Model

OpenAI has started introducing a new product to ChatGPT and API users called GPT-4o mini. This smaller language model offers enhanced textual intelligence, multimodal reasoning, and reduced usage fees for developers. Moreover, it’s positioned as the successor to GPT-3.5 Turbo.

The idea of smaller AI models isn’t new. For instance, Google has Gemini 1.5 Flash, while Anthropic has Claude 3 Haiku. These variations, including OpenAI’s, aren’t only lighter, faster, and more efficient but also more cost-effective compared to the more powerful and expensive flagship models. The latter group includes GPT-4o, Gemini 1.5 Pro, and Claude 3 Opus.

An Approach That Aims for Profitability Without Neglecting Capacity

The language models developed by OpenAI aren’t only being used in ChatGPT but are also essential components of several apps that benefit from their capabilities. However, using these models for professional purposes comes at a cost. Developers pay for groups of tokens, making fees a crucial factor to consider.

Since there’s no dominant player in this field and the competition in the AI industry has intensified, companies are striving to provide better offers to developers. We witnessed this a few months ago when OpenAI launched GPT-4o which has significantly lower usage rates compared to GPT-4 Turbo, while offering the same context length and new advantages.

Now, OpenAI has once again made a striking move by releasing GPT-4o mini, which is priced at $0.15 per million input tokens and $0.60 per million output tokens. In comparison, GPT-4o is priced at $5 per million input tokens and $15 per million output tokens. In all cases, these prices are for a 128K context window.

It’s important to consider the context window, which is the amount of information the model can use at once to generate a response. Some models have a larger context window than OpenAI’s. The Claude 3 series, for instance, has a 200K context window.

In any case, developers have numerous options both within and outside of OpenAI. The company recommends the GPT-4o mini for apps that require connections to multiple APIs, the processing of large amounts of code, or fast client interactions, such as AI-driven customer service chatbots.

On a more technical note, the GPT-4o mini currently supports text and vision in the API. OpenAI plans to incorporate text, image, video, and audio input and output support in the future. This is a significant improvement over GPT-3.5 Turbo, which was not only more expensive but also limited to text inputs and outputs.

The new smaller model achieved a score of 82% in the Measuring Massive Multitask Language Understanding (MMLU) test. Meanwhile, GPT-3.5 scored 70%, GPT-4o scored 88.7%, Claude 3 Haiku scored 75.2%, and Gemini 1.5 Flash scored 78.9%. Google Gemini Ultra claimed the top spot in the rankings, scoring 90% in this benchmark test.

Users with Free, Plus, and Team accounts for ChatGPT will soon have access to all the benefits of GPT-4o mini. The rollout has already begun. However, those with Enterprise accounts will need to wait a little longer, as GPT-4o mini will slowly roll out for them next week.

This article was written by Javier Marquez and originally published in Spanish on Xataka.

Image | OpenAI | Screenshot

Related | The EU Won’t Receive Meta’s Next Major AI Models. The Reason: Local Regulations