OpenAI has unveiled its latest small AI model, GPT-4o mini, which promises to be the most cost-efficient option available. The company anticipates that the GPT-4o mini will broaden the scope of applications built with AI by significantly reducing the cost of implementing intelligent systems.
This model is more than 60% cheaper than the previous GPT-3.5 Turbo.
According to OpenAI, the GPT-4o mini model achieves a score of 82% on the MMLU benchmark and currently outperforms GPT-4 in chat preferences on the LMSYS leaderboard. The model supports text and vision in the API, with future plans to incorporate support for text, image, video, and audio inputs and outputs.
The GPT-4o mini features a context window of 128K tokens and supports up to 16K output tokens per request. Its knowledge base is updated up to October 2023. The improved tokenizer shared with GPT-4o enhances cost-effectiveness when handling non-English text.
GPT-4o mini surpasses GPT-3.5 Turbo and other small models in academic benchmarks for both textual intelligence and multimodal reasoning. It supports the same range of languages as GPT-4o and exhibits strong performance in function calling and long-context scenarios.
On the MGSM benchmark, which measures math reasoning, GPT-4o mini scored 87.0%, outperforming Gemini Flash (75.5%) and Claude Haiku (71.7%).
In the HumanEval benchmark, which assesses coding performance, GPT-4o mini achieved a score of 87.2%, compared to Gemini Flash’s 71.5% and Claude Haiku’s 75.9%.
The GPT-4o mini model is available as a text and vision model in the Assistants API, Chat Completions API, and Batch API. Developers will be charged 15 cents per 1M input tokens and 60 cents per 1M output tokens. OpenAI plans to introduce fine-tuning for GPT-4o mini in the coming days.
In the ChatGPT application, Free, Plus, and Team users can access GPT-4o mini starting today, replacing GPT-3.5. Enterprise users will have access beginning next week. This broad availability aims to make advanced AI capabilities more accessible to a wider range of users and developers.
Conclusion
OpenAI's introduction of the GPT-4o mini model marks a significant step in making AI more affordable and accessible. With its cost-efficiency and superior performance on various benchmarks, GPT-4o mini is poised to expand the range of AI applications across different industries. As OpenAI continues to innovate and improve its models, the potential for integrating AI into everyday applications continues to grow, promising a future where intelligent systems are more widely available and cost-effective.