OpenAI slashes GPT-3.5 Turbo's cost as it prepares to ship GPT-4 Turbo with vision
What you need to know
OpenAI is unveiling a new generation of embedding models, an updated GPT-4 Turbo preview model, an updated GPT-3.5 Turbo model, and an updated text moderation model.
It's also reducing the pricing for its GPT-3.5 Turbo model and further promises better and more accurate responses.
The company is also launching a GPT-4 Turbo preview model, which will build on GPT-4's capabilities.
GPT-4 Turbo with vision is expected to ship to general availability is also expected to ship to general availability in the next few months.
OpenAI recently unveiled a new generation of embedding models, an updated GPT-4 Turbo preview model, an updated GPT-3.5 Turbo model, and an updated text moderation model. The company also indicated that it's reducing the cost of its GPT-3.5 Turbo model.
If you use ChatGPT, you've probably interacted with the GPT-3.5 Turbo model. And now, for the third time in one year, OpenAI is adjusting the pricing for this model. The company aims to help customers scale by reducing the input prices by 50%, which translates to $0.0005 /1K tokens. Output prices have also been cut by 25%, translating to $0.0015 /1K tokens.
Alongside the price reduction, OpenAI promises that the model will provide users with better and more accurate responses. The updated model also ships with a fix for a bug that previously caused a text encoding issue for non-English language function calls. The company indicated that users will automatically be upgraded to the updated GPT-3.5 model two weeks after it ships.
OpenAI is also launching an updated GPT-4 Turbo preview model, which will complement GPT -4's updated knowledge cutoff and larger 128k context windows, ultimately allowing it to handle complex tasks like code generation better and more accurately.
According to the company, the preview model is specifically designed to reduce instances of "laziness" that prevent the model from completing tasks. Moreover, it also ships with a fix for a bug affecting non-English UTF-8 generations.
OpenAI also introduces two new embedding models that promise stronger and better performance than their predecessors. Finally, GPT-4 Turbo with vision is expected to ship to general availability "in the coming months," though it's in preview.
