Downcodes editor reports: OpenAI announces big news! Its new multi-modal large-scale model GPT-4o now has fine-tuning capabilities! Developers can easily customize model behavior based on their application needs, such as adjusting the tone, setting specific instructions, or improving the accuracy of technical tasks. Even a small amount of training data can significantly improve model performance, which will greatly improve the application efficiency of the model in programming, creative writing and other fields. What’s even more surprising is that OpenAI has also launched a limited-time promotion to help developers get started quickly.
OpenAI announced that third-party developers can now fine-tune its new multi-modal large-scale model GPT-4o. This feature allows developers to adjust the model's behavior based on the needs of their application, such as changing the tone of voice, following specific instructions, or improving accuracy in technical tasks. Fine-tuning can achieve significant performance improvements even with only a small amount of training data.
For developers, using this new feature is very simple. They simply visit OpenAI's fine-tuning dashboard, click "Create," and select "gpt-4o-2024-08-06" in the base model's drop-down menu. Such convenient operation means that fine-tuning can greatly improve the performance of the model in various fields, whether it is programming or creative writing.
Entrance: https://platform.openai.com/finetune
Before September 23, 1 million Tokens will be given away for free every day
OpenAI also launched a special offer: until September 23, 2024, it will provide developers with a free usage quota of up to 1 million Tokens every day. These tokens are the basic units for the model to understand and process information. Developers need to convert data related to their own business into tokens in order to perform effective fine-tuning.
It should be noted that the usual cost of fine-tuning the GPT-4o model is US$25 per 1 million Tokens, while the cost of using the fine-tuned model during inference is US$3.75 per 1 million input Tokens, and the cost of output Tokens is US$15 per 1 million. .
For the smaller GPT-4o mini model, OpenAI provides a free quota of 2 million Tokens every day until the same deadline. Such measures not only attract a large number of developers, but also make OpenAI more competitive in the market when facing competitors such as Google and Anthropic, as well as newly launched open source models such as Nous Research's Hermes3.
In addition, while OpenAI promotes fine-tuning functions to the outside world, it does not ignore the issues of security and data privacy. They say the fine-tuned model has full control over corporate data, ensuring inputs and outputs are not used to train other models. To ensure that applications comply with OpenAI's usage policies, the company has also implemented multiple layers of security measures.
Fine-tuning success stories
There have been some successful cases of fine-tuning. For example, an AI software engineering company called Cosine achieved a top score of 43.8% on the SWE-bench benchmark through fine-tuning.
Another AI solutions company called Distyl has fine-tuned GPT-4o to achieve an execution accuracy of 71.83% in the BIRD-SQL benchmark. These successful examples undoubtedly add more convincing to OpenAI's fine-tuning capabilities.
OpenAI is working toward a vision where every organization has its own customized AI model, and this fine-tuning feature is undoubtedly an important step toward that goal.
Highlight:
** Fine-tuning function online**: OpenAI launches GPT-4o model fine-tuning function, allowing developers to adjust model behavior according to needs.
? **Free Tokens Giveaway**: Developers can receive 1 million Tokens for free every day for model fine-tuning, attracting many developers to participate.
? **Data Security Guarantee**: OpenAI attaches great importance to data privacy and security, ensuring that fine-tuned models will not be retrained with input and output data.
All in all, the fine-tuning function of OpenAI GPT-4o and its supporting free token activities provide developers with unprecedented opportunities and open a new chapter for the development of AI model customization. Act quickly and experience the powerful functions of GPT-4o!