OpenAI’s GPT-4o mini launched | Smaller and cheaper than GPT-3.5 Turbo model

GPT-4o mini has a context window of 128K tokens and supports up to 16K output tokens per request. Its knowledge cut-off is October 2023 and the model handles non-English text in a more cost-effective way

ChatGPT’s Free, Plus, and Team users can access the new model immediately [File] | Photo Credit: REUTERS

OpenAI announced the release of GPT-4o mini, which it called its “most cost-efficient small model.”

GPT-4o mini can support text and vision in the API, while support for text, image, video and audio inputs and outputs is yet to come.

Per the ChatGPT-maker, GPT-4o mini has a context window of 128K tokens and supports up to 16K output tokens per request. Its knowledge cut-off is October 2023 and the model handles non-English text in a more cost-effective way, claimed the company.

While the name might be “mini,” OpenAI stressed that the small model could hold its own against both smaller rivals as well as provide an experience comparable to larger ones.

(For top technology news of the day, subscribe to our tech newsletter Today’s Cache)

“GPT-4o mini surpasses GPT-3.5 Turbo and other small models on academic benchmarks across both textual intelligence and multimodal reasoning, and supports the same range of languages as GPT-4o,” said OpenAI.

ChatGPT’s Free, Plus, and Team users can access the new model immediately, while Enterprise users will get access from next week.

OpenAI noted that safety measures were in place from the pre-training stage so that the model would not learn from hate speech, adult content, sites that primarily aggregate personal information, and spam.

In addition, the model has been fortified to better stand against jailbreak attempts, prompt injections, and system prompt extractions.

“GPT-4o mini surpasses GPT-3.5 Turbo and other small models on academic benchmarks across both textual intelligence and multimodal reasoning, and supports the same range of languages as GPT-4o. It also demonstrates strong performance in function calling, which can enable developers to build applications that fetch data or take actions with external systems, and improved long-context performance compared to GPT-3.5 Turbo,” said OpenAI in its statement introducing the new model.

The AI company backed by Microsoft was criticised by whistleblowers and former employees who claimed that it did not take enough safety precautions when releasing new products, and that it tried to stop employees from speaking up about the same.