Sam Altman’s OpenAI startup announces a big upgrade for ChatGPT as it competes in the generative AI industry. She plans to add new text features to her powerful chatbot that will be of great use. In addition to this, it is revising its subscription offers downwards to broaden the accessibility of its product to more Internet users around the world.
New versions of GPT-3.5 Turbo and GPT-4
On June 13, 2023, OpenAI announced the release of new versions of GPT-3.5 and GPT-4, its latest generative text AI. For this major upgrade, Sam Altman’s startup is rolling out a new feature called “Function Call.” This allows developers to describe programming functions and have GPT-3.5 and GPT-4 generate the appropriate code.
As an illustration, “Function Call” can be useful for creating chatbots with multiple features. The latter will be able to answer questions by calling external tools to help users. Beyond that, they can convert natural language into database queries and extract structured data from text.
In a blog post, OpenAI clarifies that: “These models have been refined to both detect when a function should be called… and to respond with JSON that adheres to the function’s signature. The function call allows developers to more reliably retrieve structured data from the model. »
A larger popup window for GPT-3.5 Turbo
Along with the Function Call feature, OpenAI introduces in GPT-3.5 Turbo a larger popup window. This is measured in tokens or raw text chunks and refers to the text that the model considers before generating content. The new GPT-3.5 Turbo offers four times the context length (16,000 tokens) of the vanilla GPT-3.5 Turbo at twice the price.
This is approximately $0.003 for 1000 input tokens (tokens fed into the model) and $0.004 for 1000 output tokens (tokens generated by the model). OpenAI says it can generate up to 20 pages of text at once. However, this is well below the Claude flagship model from the startup Anthropic, which is capable of processing a hundred pages. OpenAI is nevertheless testing a version of GPT-4 with a context window of 32,000 tokens, but only in a limited version.
With this new policy however, the startup announces that it is reducing its costs for GPT-3.5 Turbo by 25%. Developers can now use the model of $0.0015 for 1,000 input tokens and $0.002 for 1,000 output tokens. That equates to about 700 pages per dollar, which will save them some money.
A price reduction for TEXT-embedding ada-002
During the upgrade performed by OpenAI, the startup further lowered the price for TEXT-embedding ada-002. Considered one of the most popular text embedding models on ChatGPT, they measure the relatedness of text strings. They are commonly used for search (where results are ranked by relevance to a query string) and recommendations. For the latter, elements with associated text strings are recommended.
TEXT-embedding ada-002 now costs $0.0001 for 1,000 tokens, a 75% discount from the original price. OpenAI says the reduction was made possible by increased efficiency in its systems, a key area for the startup. It also spends hundreds of millions of dollars on R&D and infrastructure.
