OpenAI Presents GPT-4o mini: Advancing Cost-Efficient

OpenAI continues its mission to democratize AI by unveiling the GPT-4o mini, a model designed to make intelligence more accessible and affordable. This new model promises to significantly broaden the scope of applications that can leverage AI, thanks to its impressive performance and reduced costs.

Revolutionary Affordability and Performance

GPT-4o mini represents a major leap in cost efficiency, priced at just 15 cents per million input tokens and 60 cents per million output tokens. This pricing makes it an order of magnitude more affordable than previous models, including the popular GPT-3.5 Turbo, which is over 60% more expensive. Despite its smaller size, GPT-4o mini excels in performance, scoring 82% on the MMLU benchmark and outperforming GPT-41 in chat preferences on the LMSYS leaderboard.

Versatile, Powerful Applications

With its low cost and latency, GPT-4o mini enables a wide range of tasks. It can handle multiple model calls in parallel, manage large volumes of context such as entire codebases or conversation histories, and provide fast, real-time text responses for applications like customer support chatbots. This versatility opens new possibilities for developers to create and scale AI-powered applications efficiently.

Multimodal Capabilities and Long-Context Handling

GPT-4o mini supports text and vision inputs and outputs, with plans to expand to video and audio in the future. It boasts a context window of 128K tokens and supports up to 16K output tokens per request. Its improved tokenizer ensures cost-effective handling of non-English text, making it a valuable tool for global applications. The model’s superior performance in both textual intelligence and multimodal reasoning sets it apart from other small models, including GPT-3.5 Turbo.

Benchmark Excellence

GPT-4o mini has been rigorously evaluated and outperforms other small models in key areas:

• Reasoning Tasks: Achieves 82.0% on MMLU, surpassing Gemini Flash and Claude Haiku.
• Math and Coding Proficiency: Scores 87.0% on MGSM and 87.2% on HumanEval, significantly outperforming competitors.
• Multimodal Reasoning: Excels with a score of 59.4% on MMMU, highlighting its robust multimodal capabilities.

Enhanced Safety Measures

OpenAI prioritizes safety, and GPT-4o mini is no exception. The model includes built-in safety features from the outset, using techniques such as reinforcement learning with human feedback (RLHF) to align its behavior with OpenAI’s policies. Extensive testing by over 70 external experts has informed improvements, making GPT-4o mini a reliable and secure choice for developers. Additionally, it is the first model to apply OpenAI’s instruction hierarchy method, enhancing its resistance to jailbreaks and prompt injections.

A Promising Future for AI Integration

The introduction of GPT-4o mini marks a significant milestone in AI development. The model’s affordability and advanced capabilities are set to pave the way for seamless AI integration in applications across various industries. OpenAI remains committed to driving down costs while enhancing model performance, ensuring that the benefits of AI become more accessible and ingrained in our daily digital experiences.

Test GPT-4o Mini in AdMind Chat

From today, it is possible to test GPT-4o mini on your AdMind Chat. This feature is available for free for all Pro account subscriptions, providing users with the opportunity to experience the model’s advanced capabilities firsthand.

In conclusion, GPT-4o mini is a testament to OpenAI’s dedication to making cutting-edge AI technology broadly available and practical for diverse applications. This new model promises to revolutionize how developers and businesses harness AI, making sophisticated intelligence more attainable and functional in real-world scenarios.