OpenAI has launched GPT-4o mini, a smaller and more cost-effective version of its AI models, aiming to make advanced AI tools more accessible and affordable for broader use. Priced at 15 cents per million input tokens and 60 cents per million output tokens, GPT-4o mini offers a significant cost reduction, over 60% cheaper than GPT-3.5 Turbo and much more affordable compared to other frontier models.
GPT-4o mini delivers strong performance across multiple domains. It achieves an 82% score on the MMLU benchmark for textual intelligence and reasoning, and excels in specific areas like math and coding, where it scored 87% on MGSM and 87.2% on HumanEval respectively. It also demonstrates solid multimodal reasoning capabilities, achieving a 59.4% score on MMMU.
This model supports a wide range of tasks due to its low cost and latency. It is capable of handling large volumes of data, such as full code bases or extensive conversation histories, and is well-suited for real-time applications like customer support chatbots. Additionally, GPT-4o mini can now process text and vision inputs within the API, with plans to support more modalities like image, video, and audio in the future.
GPT-4o mini retains the safety features of its predecessors, incorporating strategies like reinforcement learning with human feedback (RLHF) to align its outputs with safety policies. It has been rigorously tested for safety through both automated and human evaluations, ensuring reliability and ethical compliance.
Currently available through OpenAI’s Assistants API, Chat Completions API, and Batch API, GPT-4o mini is poised to enhance the efficiency and affordability of AI applications, promising a more integrated future for AI in digital environments.
For more details, visit OpenAI’s official release on GPT-4o mini.