About the new ChatGPT-4o mini

Last week, OpenAI announced their most cost-efficient small model yet GPT-4o mini, as part of their dedication to making artificial intelligence widely accessible. Priced at 15 cents per million input tokens and 60 cents per million output tokens, it's significantly cheaper than previous models, including GPT-3.5 Turbo. GPT-4o mini scores 82% on MMLU and leads GPT-41 on chat preferences in the LMSYS leaderboard.

With a context window of 128K tokens and support for up to 16K output tokens per request, this model allows for a wide range of applications, such as multi-API calls, large context processing, and real-time customer support chatbots. Currently, GPT-4o mini supports text and vision inputs and outputs, with plans to add support for images, video, and audio in the future.

?

GPT-4o mini surpasses GPT-3.5 Turbo and other small models in academic benchmarks, demonstrating superior performance in textual intelligence and multimodal reasoning. It excels in reasoning tasks, math, coding, and multimodal evaluations, outperforming models like Gemini Flash and Claude Haiku.

OpenAI tested GPT-4o mini with partners like Ramp and Superhuman to understand its capabilities and limitations. These tests showed improvements in tasks like extracting structured data and generating email responses compared to GPT-3.5 Turbo.

Safety is a priority, with built-in measures and continuous improvement based on feedback from over 70 external experts. New techniques, like the instruction hierarchy method, enhance the model's reliability and resistance to security vulnerabilities.

GPT-4o mini is available now via the Assistants API, Chat Completions API, and Batch API. Fine-tuning options will be available soon. ChatGPT Free, Plus, and Team users can access GPT-4o mini today, with Enterprise access rolling out next week.

OpenAI aims to keep reducing costs and increasing model capabilities, envisioning a future where AI seamlessly integrates into every app and website. GPT-4o mini represents a significant step towards making powerful AI applications more affordable and accessible.