OpenAI now allows businesses to fine-tune GPT-3.5 Turbo
OpenAI has announced that businesses can now custom-tune GPT-3.5 Turbo using their proprietary data. This will enable the creation of custom models that can match or even surpass GPT-4's capabilities for specific tasks. By fine-tuning, companies can develop a ChatGPT model tailored to their unique requirements, ensuring reliable responses in a particular language or with concise wording. This process also allows the AI to adopt a brand's voice for consistent advertisement copy and internal communications.
GPT-3.5 Turbo offers superior performance compared to its predecessors
Introduced earlier this year, GPT-3.5 Turbo is a versatile model family designed for applications beyond chatbots. With the ability to handle 4,000 tokens at once, it doubles the capacity of its predecessors. Early testers have reported a remarkable 90% reduction in prompt length after tuning GPT-3.5 with fine-tuned instructions. This makes it an invaluable tool for software companies, as it can efficiently handle routine coding tasks such as API calls, formatting, and completing code snippets.
How much does it cost?
GPT-3.5 Turbo's pricing is set at $0.0080 (around Rs. 0.66) for 1,000 tokens of training, $0.0120 (roughly Rs. 1) per 1,000 tokens for input usage, and $0.0120 (approximately Rs. 1) per 1,000 tokens for the chatbot's output. The model is pre-trained until September 2021 before incorporating company-specific data. OpenAI guarantees that no client data, inputs, or outputs will be utilized to train models outside of the client's organization.
GPT-4 customization on the horizon for businesses
Later this year, OpenAI intends to make GPT-4 available for custom tuning as well. This will further broaden the opportunities for businesses to develop specialized AI models that cater to their distinct needs and requirements.
