Open AI announces the launch of a new generation of embedding models that can convert text inputs into numerical forms for various machine learning applications. With the release of the new versions of GPT-4 Turbo and moderation models, the pricing of GPT-3.5 Turbo has been reduced, and the cost of accessing their models through their application programming interface (API) has been lowered. OpenAI has also announced new embedding models. In the field of artificial intelligence, embedding refers to the representation of abstract concepts in natural language or code as numerical sequences. They make it easier for machine learning algorithms to understand the relationships between these contents when performing tasks such as clustering or retrieval. It is this ability that makes them crucial for applications like knowledge retrieval, whether using ChatGPT or the Assistant API. The new embedding models include text-embedding-3-small and text-embedding-3-large. Compared to the previous text-embedding-ada-002 model, they offer developers higher performance and lower prices. More importantly, OpenAI states that they can create embeddings with up to 3,072 dimensions, which means they can capture more semantic information to improve the accuracy of downstream tasks. The company explains in a blog post that the new models help improve the average score of the most commonly used MIRACL multilingual retrieval benchmark from 31.4% to 54.9%. Additionally, the average score of the MTEB English task benchmark has increased from 61% to 64.6%. Furthermore, the pricing of text-embedding-3-small is now five times lower than text-embedding-ada-002, making it more affordable and easier for developers to access. GPT-4 Turbo and GPT-3.5 Turbo large multimodal models have also been updated to understand human text and language and generate fresh content, including code. The latest versions of these models reportedly improve instruction following, provide more reproducible outputs, and support parallel function calls. Additionally, a 16k version of GPT-3.5 Turbo has been introduced, capable of handling more complex inputs and outputs than the previous 4k version. GPT-4 Turbo has also received an interesting fix, "this model is more thorough in completing tasks like code generation compared to the previous preview models, aiming to reduce the 'laziness' of the model," wrote OpenAI researchers. OpenAI has also updated its text moderation model, which aims to detect sensitive or unsafe text inputs and outputs. According to the company, the latest version supports more languages and domains and provides better explanations for its predictions. Finally, OpenAI is launching a new way for developers to manage their API keys and understand their usage of the service. With this update, developers can create multiple API keys for each application or service and monitor their individual usage and billing details within the OpenAI dashboard. This helps them better understand the costs of embedding OpenAI models for each application or service. Additionally, the company states that it will reduce the price of accessing GPT-3.5 Turbo through the API by 25%, making it more accessible. According to OpenAI, these updates are part of their ongoing efforts to improve the quality and capabilities of their AI models while making them more useful and cost-effective for developers. The company also invites developers to participate in evaluations to help improve model capabilities for different workloads.