GPT-4o, OpenAI’s latest and most powerful artificial intelligence (AI) model released in May, is getting a new upgrade. On Tuesday, the company released a new fine-tuning feature for the AI model that allows developers and organizations to train it using custom datasets. This will allow users to add more relevant and targeted data related to their usage and make the responses generated more accurate. For the coming month, the AI company also announced that it will provide free training tokens to organizations to improve the GPT-4o models.
GPT-4o gets a fine-tuning function
In a afterOpenAI announced the launch of the new feature, emphasizing that it will enable developers and organizations to get higher performance at a lower cost for specific use cases. The AI company called it “one of the most requested features from developers,” explaining that fine-tuning the model will allow it to adjust the structure and tone of responses. It will also allow GPT-4o to follow complex domain-specific instructions.
In addition, the company also announced that it will provide organizations with free training tokens for its AI models through September 23. Companies using GPT-4o will receive one million training tokens per day, while companies using GPT-4o mini will receive two million training tokens per day.
Additionally, fine-tuning the training of the models costs $25 (roughly Rs. 2,000) per million tokens. Further, inference costs $3.75 (roughly Rs. 314) per million input tokens and $15 (roughly Rs. 1,250) per million output tokens, OpenAI said.
To fine-tune GPT-4o, users can go to the fine-tuning dashboard that will open in a new window, click Create and select “gpt-4o-2024-08-06” from the base model dropdown. To do the same for the mini model, users should select the base model “gpt-4o-mini-2024-07-18”. These AI models are only available to developers who subscribe to OpenAI’s paid tiers.
Fine-tuning in this context is essentially a method of obtaining the full processing capabilities of the large language model (LLM) while curating specific datasets to make it more amenable to niche workflows. It works like an AI agent or GPTs, but it is not limited in processing power, resulting in faster and generally more accurate responses.
