In a significant update, OpenAI has enabled fine-tuning capabilities for its lightweight GPT-3.5 Turbo model. The new feature allows customers to improve the model's performance, enabling it to excel in narrow tasks that could be comparable to or even better than OpenAI's flagship GPT-4 model.
The fine-tuning feature comes as a response to growing demands from developers and businesses seeking to customize the AI model for unique and specific user experiences. With fine-tuning, GPT-3.5 Turbo can be modified to follow detailed instructions, format responses consistently, and adapt its tone to match a particular brand or voice.
Fine-tuning involves uploading data files and creating a fine-tuning job via OpenAI's API. The process goes through a moderation system, which checks for compliance with OpenAI's safety standards. The company plans to introduce a fine-tuning UI with a dashboard to track the status of ongoing workloads.
Fine-tuning costs are structured as follows:
- Training: $0.008 per 1,000 tokens
- Usage input: $0.012 per 1,000 tokens
- Usage output: $0.016 per 1,000 tokens
For example, a fine-tuning job with a training file of 100,000 tokens would cost approximately $2.40.
OpenAI also announced updates to two GPT-3 base models—babbage-002 and davinci-002—which are now fine-tunabale and offer more extensibility. The original GPT-3 base models will be retired in January 2024.
OpenAI has revealed that GPT-4, which can understand both text and images, will also receive fine-tuning capabilities sometime later this fall.
- Enhanced Customization: GPT-3.5 Turbo can now be fine-tuned for specialized tasks.
- Cost Efficiency: Fine-tuning allows for shorter text prompts, potentially reducing API costs.
- Safety Checks: All fine-tuning data goes through a GPT-4 powered moderation system to ensure compliance with OpenAI's safety standards.
- Upcoming UI: OpenAI plans to introduce a dashboard for easier management of fine-tuning tasks.
The new fine-tuning feature for GPT-3.5 Turbo offers companies the ability to create more specialized and reliable AI models. With promising applications and competitive pricing, it marks a significant step in customizable AI technology.
Stay tuned for further updates on OpenAI's innovative enhancements to its AI models.
