OpenAI releases new models and lowers API pricing



OpenAI has recently unveiled a series of enhancements aimed at benefiting developers utilizing its AI services. The updates encompass the introduction of new embedding models, a reduced price for GPT-3.5 Turbo, an enhanced preview of GPT-4 Turbo, and more robust content moderation capabilities.

In terms of text embedding models, the newly introduced text-embedding-3-small and text-embedding-3-large models exhibit improved performance compared to their predecessors. Notably, text-embedding-3-large has demonstrated average scores of 54.9 percent on the MIRACL benchmark and 64.6 percent on the MTEB benchmark, surpassing the older text-embedding-ada-002 model, which scored 31.4 percent and 61 percent, respectively.

OpenAI has also implemented a significant price reduction for text-embedding-3-small, decreasing the cost per 1,000 tokens from $0.0001 to $0.00002, making it more cost-effective for developers. Additionally, developers have the flexibility to truncate embeddings, further reducing costs without compromising accuracy significantly.

Scheduled for release next week, the updated GPT-3.5 Turbo model will see a 50 percent reduction in pricing for input tokens and a 25 percent reduction for output tokens. This marks the third price reduction for GPT-3.5 Turbo within the past year, aligning with OpenAI’s goal to encourage wider adoption.

The GPT-4 Turbo preview has also received an update to version gpt-4-0125-preview, with notable improvements in tasks such as code generation. Over 70 percent of requests have transitioned to this model since its initial release.

In the realm of content moderation, OpenAI has introduced text-moderation-007, its most advanced model to date. This model is designed to identify potentially harmful text with greater accuracy than previous iterations.

Furthermore, OpenAI has enhanced developer control by providing more options for API keys. Developers can now assign permissions to keys and gain visibility into usage metrics on a per-key basis, facilitating better tracking of individual products or projects.

OpenAI underscores its commitment to ongoing improvements, with plans for more enhancements in the coming months to better accommodate larger development teams.

Other news

US Chief Justice: AI won’t replace judges but will ‘transform our work’ – 03.01.2024

In the Federal Judiciary’s year-end report, US Chief Justice John Roberts addressed the potential impact of AI on the judicial system. In particular, he aimed to quell concerns about the obsolescence of judges in the face of technological advancements.

Read More

Google launches Gemini to replace Bard chatbot

Google’s initial AI, Bard, aimed to compete with chatbots like ChatGPT but faced criticism for underwhelming demos, leading to a rebranding as Gemini. Now, Gemini Advanced, featuring Ultra 1.0, emerges as Google’s premier model, impressing third-party evaluators in tasks like coding and logical reasoning.

Read More