OpenAI Launches Structured Outputs for JSON in Its API, Reduces GPT-4o Prices
OpenAI has introduced Structured Outputs in the API and has also reduced the costs of GPT-4o. Learn more about the developments and their implications for OpenAI customers.
- OpenAI has announced that the new GPT-4o model is now for a reduced price of $2.50 per 1 million input tokens and $10.00 per 1 million output tokens.
- OpenAI has also updated its API with support for Structured Outputs to ensure that the outputs match the JSON Schemas provided by developers.
In a development likely to be viewed positively by developers, OpenAI has introduced support for Structured Outputs for JSON in its API. This enables strict compliance with JSON Schemas for better reliability of LLM (large language model) outputs. The update also bolsters the capabilities of GPT-4o models, especially when working with structured data extraction or complex workflows. The new update complements JSON mode, which was launched last year.
The new API supports function calling and a response format using JSON Schema. In addition, the AI startup also announced a significant price reduction for GPT-4o to better compete against AI service providers like Google and Anthropic. This includes 50% lower input costs ($2.50 per million tokens), 33% lower output costs ($10.00 per million tokens), and 50% overall cost reduction compared to the previous GPT-4o model.
See More: Interpol Disrupts Massive Business Email Scam, Recovers Over $40 Million
Developer Benefits
According to OpenAI, the new GPT-4o models, such as GPT-4o-2024-08-06, have shown superior reliability in matching JSON Schema in internal evaluations. This is a notable improvement over previous models. The new model introduces native SDK (software development kit) support for Structured Outputs in Node.js and Python. This will make it substantially easier for developers to utilize and integrate these features in applications. Structured Outputs also allow models to detect and handle unsafe requests automatically.
Some limitations include allowing only a subset of JSON Schema, additional latency in first API responses, token limits, and lack of compatibility with parallel function calls. The new feature will be available in all models that support function calling, including GPT-4o, GPT-4o-mini, GPT-4-0613, and GPT-3.5-Turbo-0613.
The new feature is essential for OpenAI as it faces competition from Google and other AI firms in the market. In addition, the price reduction is expected to make GPT-4o more accessible and will attract more organizations and developers to OpenAI platforms.
Takeaways
The introduction of Structured Outputs fixes a fundamental problem for developers, especially in environments requiring higher levels of precision, such as the development of multi-step workflows and data entry. It also improves the reliability of OpenAI LLMs and simplifies processes for developing complex applications requiring strict data conformity. As competition in the AI industry continues to intensify, such strategic updates by OpenAI in features and pricing will solidify its position in the market in the coming months.
LATEST NEWS STORIES
