OpenAI Introduces GPT-4o Mini: A New Era in Compact and Cost-Efficient AI Models Indul Hassan, July 19, 2024 In a groundbreaking move, OpenAI has taken a significant step forward in making artificial intelligence more accessible and affordable with the launch of GPT-4o Mini. This new model promises to revolutionise the AI landscape by delivering powerful capabilities at a fraction of the cost of previous models. Here’s a closer look at what makes GPT-4o Mini a game-changer. Superior Performance at Unmatched Cost-Efficiency GPT-4o Mini is designed to expand the range of AI applications by offering intelligence at a significantly lower price point. Priced at 15 cents per million input tokens and 60 cents per million output tokens, GPT-4o Mini is an order of magnitude more affordable than earlier frontier models and over 60% cheaper than GPT-3.5 Turbo. High Benchmark Scores and Advanced Capabilities The model excels in several key areas: Textual Intelligence and Reasoning: GPT-4o Mini scores 82% on the MMLU benchmark, outperforming competitors like Gemini Flash and Claude Haiku. Mathematical and Coding Proficiency: It achieves 87% on MGSM for math reasoning and 87.2% on HumanEval for coding, again surpassing other small models. Multimodal Reasoning: With a score of 59.4% on the MMMU benchmark, GPT-4o Mini demonstrates strong performance in tasks involving both text and vision. Versatile and Future-Ready Currently, GPT-4o Mini supports text and vision inputs via the API, with plans to extend support to image, video, and audio inputs and outputs. It features a context window of 128K tokens and supports up to 16K output tokens per request, with knowledge up to October 2023. Enhanced handling of non-English text further broadens its application scope. Robust Safety Measures OpenAI has integrated comprehensive safety protocols into GPT-4o Mini, filtering out harmful content during pre-training and aligning model behavior with policies using reinforcement learning with human feedback (RLHF). The model also employs a new instruction hierarchy method to resist jailbreaks, prompt injections, and system prompt extractions, making it safer for large-scale use. Proven in Real-World Applications Partners such as Ramp and Superhuman have tested GPT-4o Mini, finding it significantly better than GPT-3.5 Turbo for tasks like extracting structured data from receipts and generating high-quality email responses. Broad Availability and Accessibility GPT-4o Mini is available through the Assistants API, Chat Completions API, and Batch API. It is accessible to Free, Plus, and Team users of ChatGPT starting today, with Enterprise access rolling out next week. Fine-tuning capabilities will be introduced shortly. Model Evaluation Scores Impact on the AI Ecosystem The introduction of GPT-4o Mini is set to have a profound impact on the AI ecosystem. By lowering the barriers to entry, OpenAI is democratising access to advanced AI technologies. This could lead to a surge in innovation as more developers experiment with and build upon the capabilities of GPT-4o Mini. Furthermore, GPT-4o Mini’s efficiency and affordability are likely to drive widespread adoption in sectors that have previously been hesitant to invest in AI due to cost or technical constraints. Education, healthcare, and non-profits, in particular, stand to benefit significantly from this development. Looking Ahead OpenAI envisions a future where AI models are seamlessly integrated into every app and website, driving down costs while enhancing capabilities. With GPT-4o Mini, the future of AI is becoming more accessible, reliable, and embedded in our daily digital experiences enabling developers to build and scale powerful AI applications more efficiently than ever before. AI Generative AI aigenerativeaillm