In a strategic move to elevate the capabilities of its language models and cater to developers' evolving needs, OpenAI has announced a sweeping array of updates, including the launch of new embedding models, an enhanced GPT-4 Turbo, a refined GPT-3.5 Turbo model with reduced prices, an upgraded moderation model, and innovative API management tools.
OpenAI is introducing two cutting-edge embedding models designed to meet diverse requirements. The first, text-embedding-3-small, is a highly efficient model that outperforms its predecessor, the text-embedding-ada-002, across various benchmarks. With a 5X reduction in pricing compared to its forerunner, text-embedding-3-small provides developers with a cost-effective solution without compromising performance.
The second, text-embedding-3-large, represents the next generation of larger embedding models, boasting up to 3072 dimensions. This model exhibits superior performance, with benchmark scores surpassing the previous text-embedding-ada-002. Priced at $0.00013 per 1k tokens, text-embedding-3-large offers enhanced capabilities for applications requiring more significant embeddings.
Both models support a unique feature that allows developers to shorten embeddings without sacrificing their concept-representing properties. By utilizing the dimensions API parameter, developers can tailor embeddings to their specific needs, optimizing for performance and cost.
OpenAI is set to launch a new GPT-3.5 Turbo model, gpt-3.5-turbo-0125, accompanied by a significant price reduction. Input prices for the model have been slashed by 50% to $0.0005 per 1K tokens, while output prices are reduced by 25% to $0.0015 per 1K tokens. This move is aimed at empowering developers to scale their projects more affordably. The new model promises improvements in responsiveness and accuracy in requested formats and resolves a bug affecting non-English language function calls.
To ensure a seamless transition, customers using the pinned gpt-3.5-turbo model alias will be automatically upgraded from gpt-3.5-turbo-0613 to gpt-3.5-turbo-0125 two weeks after the model's launch.
Over 70% of requests from GPT-4 API customers have successfully transitioned to GPT-4 Turbo since its release, leveraging its updated knowledge cutoff, more oversized context windows, and lower prices. OpenAI is now introducing an updated GPT-4 Turbo preview model, gpt-4-0125-preview. This model excels in tasks like code generation and aims to mitigate cases of incomplete task completion, addressing concerns of model "laziness." The release includes a much-needed fix for a bug impacting non-English UTF-8 generations. To simplify access to the latest GPT-4 Turbo preview versions, a new gpt-4-turbo-preview model name alias has been introduced.
As part of OpenAI's ongoing commitment to safety, the free Moderation API now includes text-moderation-007, the most robust moderation model to date. With improvements in identifying potentially harmful text, the text-moderation-latest and text-moderation-stable aliases now point to this advanced model.
OpenAI is rolling out two platform improvements to empower developers with more visibility and control over API keys. Firstly, developers can now assign permissions to API keys directly from the API keys page, offering flexibility in access levels for different use cases. Secondly, the usage dashboard and export function now provide metrics on an API-critical level after turning on tracking, allowing users to monitor usage patterns more effectively.
In the coming months, OpenAI plans to further enhance features for developers to view API usage and manage API keys, mainly focusing on the needs of larger organizations.
OpenAI's latest updates mark a significant leap forward, not only in the sophistication of its models but also in addressing the practical needs of developers. Introducing new embedding models, refining existing language models, and implementing advanced API management tools reflect OpenAI's commitment to providing cutting-edge solutions in the rapidly evolving landscape of AI and natural language processing. As these updates come into effect, developers can anticipate a more streamlined and cost-effective experience, unlocking new possibilities for innovation and application development.