OpenAI Unveils Upgrades and Price Cuts Across AI Models

Key Points

• New text-embedding-3-small and text-embedding-3-large models
• Lowered pricing for GPT-3.5 Turbo with upcoming GPT-4 Turbo preview
• Enhanced content moderation capabilities with text-moderation-007

OpenAI, the renowned AI lab, has rolled out a series of updates aimed at empowering developers leveraging its AI services.

Among the highlights are the introduction of new embedding models and a significant price reduction for GPT-3.5 Turbo, along with a glimpse of the upcoming GPT-4 Turbo preview and strengthened content moderation capabilities.

The newly introduced text-embedding-3-small and text-embedding-3-large models boast superior performance compared to their predecessors. For instance, text-embedding-3-large showcases average scores of 54.9 percent on the MIRACL benchmark and 64.6 percent on the MTEB benchmark. This represents a notable improvement from the older text-embedding-ada-002 model, which scored 31.4 percent and 61 percent on the same benchmarks, respectively.

In a move to make AI services more accessible, OpenAI has slashed the price per 1,000 tokens by 5x for text-embedding-3-small when compared to text-embedding-ada-002, dropping from $0.0001 to $0.00002. Developers can further optimize costs by shortening embeddings without a significant impact on accuracy.

Looking ahead, OpenAI is set to release an updated GPT-3.5 Turbo model next week, accompanied by a 50 percent reduction in pricing for input tokens and a 25 percent reduction for output tokens. This marks the third price cut for GPT-3.5 Turbo in the past year, underscoring OpenAI's commitment to fostering wider adoption of its AI models.

The GPT-4 Turbo preview has also received an update to version gpt-4-0125-preview, witnessing a transition of over 70 percent of requests to the model since its initial launch. Improvements include more thorough code generation and enhanced performance across various tasks.

In the realm of safety, OpenAI introduces its most advanced content moderation model yet, text-moderation-007. This model excels in identifying potentially harmful text with higher accuracy than its predecessors, contributing to a safer AI environment.

For developers, OpenAI has extended enhanced control over API keys and improved visibility into usage metrics. Developers can now assign permissions to keys and gain insights into consumption on a per-key level, offering a more nuanced tracking mechanism for individual products or projects.

OpenAI has underlined that more platform improvements are on the horizon in the coming months, catering to the evolving needs of larger development teams.