Open AI Unveils GPT-4 Turbo with Improved Responsiveness and Price Reductions for GPT-3.5 Turbo
Open AI Unveils GPT-4 Turbo with Improved Responsiveness and Price Reductions for GPT-3.5 Turbo: OpenAI recently shared on its blog several enhancements concerning its text embedding and moderation models, along with updates to its flagship models, GPT-4 Turbo and GPT-3.5 Turbo. These innovations aim to provide developers with increased performance, efficiency, customization, and security for their natural language or code-based applications.
The initial version of GPT-4 was released by OpenAI in March 2023, made available to all developers in July of the same year. During its first developer conference, OpenAI DevDay, on November 6, the company unveiled a preview of the next generation of this model: GPT-4 Turbo.
Various users reported issues with ChatGPT, stating that it refused to respond to their queries or did so only partially, prompting them to complete the responses.
On December 8, OpenAI posted on X:
"We have heard all your feedback about GPT-4 becoming lazier! We have not updated the model since November 11, and it is certainly not intentional. The model's behavior can be unpredictable, and we are working to address it."
Therefore, on January 25, OpenAI released an updated preview model of GPT-4 Turbo, named "gpt-4-0125-preview," indicating that "this model performs tasks such as code generation more thoroughly than the previous preview model and is intended to reduce cases of 'laziness' when the model fails to perform a task."
Bugs resulting from difficulties encountered by the previous preview model in generating text in languages other than English have also been addressed.
GPT-3.5 Turbo Model Update OpenAI has also announced the release of a new GPT-3.5 Turbo model this week, with reduced prices (query cost halved) and performance improvements. This update aims to provide increased accuracy in responses and address coding issues encountered in the previous version, offering an enhanced experience for users.
Two New Embedding Models at Lower Prices An embedding is a sequence of numbers representing the concepts of content such as natural language or code. Embeddings enable machine learning models and other algorithms to easily understand relationships between content and perform tasks such as clustering or searching. They power applications like knowledge retrieval in ChatGPT and the API Assistants, as well as various augmented generation retrieval (RAG) development tools.
OpenAI introduces two new text embedding models: a smaller and highly efficient model, embedding-3-small, and a larger and more powerful model, embedding-3-large. These models, more efficient and cost-effective than their predecessor, embedding-ada-002, provide developers with the option to choose the embedding size based on the specific needs of their applications. This increased flexibility allows developers to optimize performance while managing the costs associated with using AI models.
Improved API Key Management Developers now benefit from new API key management features, allowing them to assign specific permissions and track API usage in more detail. These enhancements offer better control and visibility over the use of AI resources, facilitating project and budget management.
OpenAI specifies that in the coming months, it plans to further improve developers' ability to monitor their API usage and manage API keys, particularly in large enterprises.
Moderation Model Update OpenAI introduces a new, more robust moderation model that can identify potentially harmful texts with greater accuracy. According to OpenAI, this update reflects its commitment to the safety and reliability of its products, ensuring a safe and positive user experience.