OpenAI has recently unveiled GPT-4 Turbo, the latest version of its powerful AI model designed for API services. This upgraded version boasts various improvements to rival its competitors, including support for large-scale input of 128k tokens, surpassing Anthropic’s 100k token support. Additionally, the model’s training with new knowledge has enabled it to predict events up until April 2023. Other notable enhancements include:
– Ability to make multiple API calls, compared to the previous single call limitation.
– Introduction of a JSON mode, providing responses exclusively in JSON format.
– Replication of results by using the seed parameter, ensuring consistent output when keeping the seed value the same.
– Support for image input, allowing users to provide image URLs or directly embed base64 images in the API.
Moreover, GPT-4 Turbo offers a more affordable pricing structure compared to its predecessor, GPT-4, with input costs reduced to one-third and output costs halved. The pricing for image usage is calculated based on image size.
In addition to the new model, OpenAI has also announced the Assistants API, which includes a crucial feature – the ability to retain contextual information without limitations. Previously, developers had to manually input the original text for API calls. But with this update, the Assistants API can store contextual information and read documents without the need for developers to create their own vector databases. Furthermore, this API includes an embedded Python Interpreter for data processing and supports external function calls.
Usage costs for the Assistant API are calculated based on the amount of data input, at a rate of $0.20 per gigabyte per day. Running Python scripts within the API session incurs an additional cost of $0.03 per execution.
In summary, OpenAI’s release of GPT-4 Turbo and the Assistants API brings significant advancements in AI language models and API services, offering enhanced capabilities and more cost-effective solutions for a wide range of applications.
TLDR: OpenAI introduces GPT-4 Turbo, a superior AI model with expanded features and performance rivaling other models. This latest version supports large input sizes, introduces a JSON response mode, allows result replication, and accepts image input. The pricing is more economical than GPT-4, and the Assistants API offers features like unlimited memory, embedded Python Interpreter, and external function calls. Overall, OpenAI’s advancements ensure better language comprehension and more efficient AI integration.