OpenAI is intensifying its price war with other GenAI LLMs, especially Google’s Gemini. The company is doing this by further lowering the cost of GPT-4o. In addition, the AI giant is making it possible for JSON outputs from its LLMs to 100 percent match developers’ JSON Schemas.
Competition among LLMs is cutthroat, and OpenAI must also respond. Recently, the AI giant announced that it will lower the prices of its latest LLM GPT-4o.
This latest version, fully named gpt-4o-2024-08-06, now costs 2.50 dollars (2.29 euros) for the input of 1 million tokens and 10 dollars (9.16 euros) for the output of 1 million tokens. Compared to the previous version, this is 50 percent cheaper for input tokens. For output tokens, the price reduction amounts to one-third.
By comparison, using Google’s Gemini 1.5 Pro LLM costs 3.50 dollars (3.21 euros) per 1 million input tokens and 10.50 dollars (9,62 euros) per 1 million output tokens.
Structured Outputs
Not only is OpenAI lowering prices, but it is also now introducing Structured Outputs in its API. This feature ensures that valid JSON outputs generated via JSON mode from its GPT-LLMs exactly match developers’ JSON Schemas.
Availability
To this end, the company has provided its Python and Node SDKs with native support for Structured Outputs. The functionality is now available to all OpenAI LLMs that support function calling in the API. These include gpt-4o, gpt-4o mini, gpt-4-0613 and gpt-3.5-turbo-0613, as well as more finetuned models.
Structured Outputs with so-called response formats are also available for gpt-4o-mini, gpt-4o-2024-08-06 and all finetuned models based on these LLMs.
For example, the feature is suitable for the Chat Completions API, the Assistants API, and the Batch API. OpenAI indicates that it even works with vision inputs (i.e., image analysis and generating text and explanations based on that).
However, there are still some limitations. For example, JSON Schemas with Structured Outputs are not possible for Zero Data Retention.
Yet another co-founder leaves OpenAI
The battle between LLM providers also sometimes results in personnel transfers. For example, co-founder John Schulman has traded OpenAI for competitor Anthropic. According to Schulman on X, this explicitly is not because of a dispute over the company’s direction but because of personal ambitions. Another OpenAI co-founder, Ilya Sutskever, stepped down from the role of Chief Scientist some time ago, a role he had held since 2018.
CEO Sam Altman called his former colleague Sutskever a “genuinely remarkable genius” at the time. As friendly as the OpenAI boss sounded, Sutskever cleared the field a few months after he and a few others fired Altman. That decision was reversed during a remarkable soap opera in which Microsoft claimed a board role for itself, albeit without voting rights.
Also read: OpenAI’s business model isn’t working as bankruptcy looms