OpenAI announces updated GPT models for developers; reduces pricing
The AI arms race is getting fierce day by day. OpenAI has been leading the race since the beginning, and it is in no mood to lose the lead to its competitors. The company has now announced updated versions of its GPT-3.5-turbo and GPT-4 API models for developers. OpenAI has also reduced the pricing of its models to compete with its rivals.
Why does this story matter?
OpenAI started the ongoing AI revolution with the release of ChatGPT. The company took one step further when it announced GPT-4. Despite its competitive edge, the AI start-up is facing stiff competition from the likes of Google, Anthropic, and Amazon. With CEO Sam Altman confirming OpenAI hasn't started training GPT-4's successor, updating the existing models is essential for the company.
Updated GPT models have 'function calling' capability
OpenAI announced the new API updates through a blog post. The updated versions of GPT-3.5-turbo and GPT-4 come with a capability called 'function calling.' Function calling allows developers to describe functions to the models and the model can intelligently create a JSON object to execute the functions. According to OpenAI, this allows developers to more reliably connect GPT's capabilities "with external tools and API."
How will function calling help developers?
In its blog post, OpenAI has described how function calling will help developers. For instance, they can create chatbots that answer questions by calling external tools, convert natural language into API calls or database queries, or extract structured data from text. The models are "fine-tuned" to detect "when a function needs to be called" and respond with code that "adheres to the function signature."
OpenAI expanded the context window of GPT-3.5-turbo
The API updates are not limited to function calling. OpenAI has also made GPT-3.5-turbo more controllable with system messages, helping developers shape the behavior of the model. The company has also updated the context window of GPT-3.5-turbo. The new GPT-3.5-turbo with an expanded context window offers 16,000 tokens, which is four times the vanilla model.
GPT-4 with a 32,000-token context window is being tested
A model with an expanded context window will consider more of the previously generated text before generating additional text. The new GPT-3.5-turbo will be able to consider around 20 pages in one go. The company is also testing a 32,000-token context window for GPT-4. Meanwhile, OpenAI rival Anthropic's Claude has a context window of 100,000 tokens.
OpenAI cuts vanilla GPT-3.5-turbo's input token cost by 25%
OpenAI has reduced the pricing of its API models. The input tokens of the original GPT-3.5-turbo now cost 25% less. The model can be used for $0.0015 per 1,000 input tokens and $0.002 per 1,000 output tokens, which comes to around 700 pages per $1. The start-up also cut the costs of text-embedding-ada-002, its popular embedding model, by 75% to $0.0001 per 1,000 tokens.