OpenAI Reduces Model Costs to Enhance Accessibility

OpenAI has recently introduced a range of exciting updates, including new embedding models and significant price reductions for its popular GPT-3.5 Turbo model. For the third time this year, the input costs for GPT-3.5 Turbo have been slashed by 50%, bringing the price down to just $0.0005 per 1,000 tokens. Additionally, output costs have been lowered by 25% to $0.0015 per 1,000 tokens. This adjustment positions GPT-3.5 Turbo as a more economical option compared to Anthropic’s Claude 2.0 and 2.1, which charge $0.008 per 1,000 tokens for inputs and $0.024 for outputs.

In conjunction with the pricing updates, OpenAI is launching an enhanced version of GPT-3.5 Turbo, designated as gpt-3.5-turbo-0125. This revised model features improved accuracy in responding to specified formats and addresses a bug related to text encoding in non-English language function calls. Furthermore, OpenAI has rolled out an advanced preview of GPT-4 Turbo, known as gpt-4-0125-preview. This model excels in task completion, offers a larger context window for inputs—approximately 85,333 words—and comes with reduced pricing.

Looking ahead, OpenAI has announced the upcoming general availability of a GPT-4 Turbo version equipped with vision capabilities, expected to make its debut in the coming months.

### New Embedding Models

Accompanying the model upgrades, OpenAI has unveiled a set of new embedding models designed to enhance data processing. Embedding models convert high-dimensional data, such as text or images, into a lower-dimensional space, facilitating easier computation of similarities between data points.

The new text-embedding-3-small model serves as an upgrade to its predecessor, text-embedding-ada-002, which was released in December 2022. This latest model boasts improved performance in multilingual retrieval and English language tasks and is now five times cheaper, priced at just $0.00002 per 1,000 tokens.

In addition, OpenAI introduced a large-scale embedding model called text-embedding-3-large, which offers enhanced performance for more demanding applications, with a cost of $0.00013 per 1,000 tokens. Both models have been developed using innovative techniques that allow developers to optimize performance and cost by shortening embeddings without losing essential conceptual representation.

### New Moderation Model

To further bolster its commitment to safety, OpenAI has released text-moderation-007, a new moderation model designed to identify potentially harmful text. This release comes as part of the organization's continuous effort to reinforce safety measures, particularly following a recent leadership transition. The new moderation model allows developers to ensure their content aligns with OpenAI's usage policies and effectively detects potentially problematic outputs.

These updates reflect OpenAI’s intention to provide users with more powerful tools at reduced costs, enhancing the overall capabilities of its AI offerings while maintaining a strong focus on safety and performance.