AI BriefingGPT-4o Fine-tuning Feature Launched
Key point
OpenAI has launched a GPT-4o fine-tuning feature that lets developers optimize the model for specific use cases.
Details
OpenAI has launched the fine-tuning feature for GPT-4o, which developers had been requesting a lot. Developers can use custom datasets to adjust the model's response structure and tone, or optimize the model to follow complex domain-specific instructions.
It is currently available to developers on all paid usage tiers, and GPT-4o mini fine-tuning is also offered together. As part of a promotion, until September 23, all organizations will be given 1 million free training tokens per day for GPT-4o and 2 million free training tokens per day for GPT-4o mini.
Key achievements and security details are as follows:
- Cosine: Using fine-tuned GPT-4o, achieved a SOTA (State-of-the-art) record on the SWE-bench benchmark.
- Distyl: Ranked #1 on the BIRD-SQL benchmark, demonstrating excellent SQL generation capability.
- Data security: Fine-tuned models remain under the user's full control, and input and output data are not used to train other models.
The cost of GPT-4o fine-tuning is $25 per 1 million tokens for training, and for inference, $3.75 per 1 million input tokens and $15 per 1 million output tokens.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.