OpenAI has released a new version of GPT-4 Turbo and reduced API usage fees, adding multimodal capabilities and developer tools.

On November 7th, at the OpenAI DevDay event, OpenAI announced dozens of new additions and improvements, as well as implemented price reduction strategies in multiple parts of the platform. These updates include the launch of a new, more powerful and cost-effective GPT-4 Turbo model that supports a 128K context window, a new Assistants API that facilitates developers in building their own AI applications, and the addition of new multimodal features in the platform, including vision, image creation (DALL·E 3), and text-to-speech (TTS). These new features will be rolled out to OpenAI customers starting from 1 pm (Pacific Time) today. The GPT-4 Turbo model not only outperforms its predecessors in processing capabilities but also extends its understanding of world events up to April 2023 and supports a larger context window, capable of processing the information equivalent to over 300 pages of text in a single prompt. Additionally, OpenAI has optimized its performance and is able to provide input tokens at one-third the price of GPT-4, and the output token price has also been reduced by half. In addition, OpenAI announced updates to function calls, improved compliance with instructions and JSON schemas, introduced new features for reproducible outputs and log probabilities, and updated the GPT-3.5 Turbo model. OpenAI also launched new tools such as Assistants API, retrieval, and code interpreter, aimed at helping developers build higher quality AI applications. These APIs are designed to be flexible and adaptable to various use cases, from natural language data analysis applications and coding assistants to AI-driven holiday planners, voice-controlled DJs, and even intelligent visual canvases. The new APIs also bring new multimodal capabilities, such as the ability for the GPT-4 Turbo model to accept images as inputs in the Chat Completions API and support integration with DALL·E 3 through the API, as well as a brand new text-to-speech (TTS) model. Additionally, OpenAI announced a series of price reductions and higher rate limits, as well as enhanced copyright protection measures. OpenAI also mentioned that they are creating an experimental access program for GPT-4 fine-tuning and launching a custom model program for organizations that require customization beyond fine-tuning. OpenAI provides an application link for organizations interested in participating.