GPT-4o mini: OpenAI Launches Intelligent and Economical Small Model

OpenAI has introduced the new GPT-4o mini model, an intelligent, economical, and fast small model that is smarter and cheaper than GPT-3.5 Turbo while maintaining the same speed. This article will detail the features and advantages of this new model.

from OpenAI GPT-4o mini

Introduction

OpenAI today(7/18) announced the launch of GPT-4o mini, a brand new artificial intelligence model designed to provide users with a smarter and more economical option. GPT-4o mini not only outperforms GPT-3.5 Turbo in performance but also significantly reduces usage costs, bringing new possibilities for developers and businesses.

Differences from GPT-3.5

1. Intelligence Level: GPT-4o mini excels in textual intelligence, scoring 82% on the MMLU test, far surpassing GPT-3.5 Turbo’s 69.8%.
2. Price Advantage: GPT-4o mini is over 60% cheaper than GPT-3.5 Turbo, costing only $0.15 per million input tokens and $0.60 per million output tokens.
3. Multilingual Ability: GPT-4o mini shows significant improvement in understanding non-English languages.
4. Context Window: GPT-4o mini has a 128k context window, supports up to 16k output tokens, and has a knowledge cut-off date of October 2023.

Features and Characteristics

1. Multimodal Capabilities: Currently supports text and vision capabilities, with plans to add audio and video input and output support in the future.
2. Applicable Scenarios:
   • High-volume tasks: Such as passing a full code base or conversation history.
   • Cost-sensitive tasks: For example, summarizing large documents.
   • Tasks requiring fast responses: Such as customer support chatbots.
3. Batch Processing Discount: 50% discount available when using the Batch API for batch jobs completed asynchronously within 24 hours.
4. Upcoming Features: OpenAI plans to launch fine-tuning for GPT-4o mini in the coming days.

Conclusion

The launch of GPT-4o mini provides developers and businesses with a smarter and more economical choice. It not only outperforms GPT-3.5 Turbo in performance but also significantly reduces usage costs. OpenAI recommends that developers currently using GPT-3.5 Turbo switch to GPT-4o mini to achieve higher intelligence levels at a lower cost. Whether in the Chat Completions API, Assistants API, or Batch API, GPT-4o mini will bring more value and possibilities to users.