GPT-4.5 Turbo Details Leak Through Search Engines Ahead of Rumored OpenAI Announcement

GPT-4.5-Turbo seems to be indexed on Bing… coming really soon? pic.twitter.com/lzZpQzkoSS

— Matt Shumer (@mattshumer_) March 12, 2024

Details about OpenAI’s latest AI model, GPT-4.5 Turbo, have been prematurely leaked through search engine indexing by Bing and DuckDuckGo. The product page for GPT-4.5 Turbo has appeared in search results ahead of an official release, leading to speculation about the features and capabilities of OpenAI’s newest model.

GPT-4.5 Turbo

Although the link provided in the search results currently redirects to a 404 error page, snippets from the page suggest that GPT-4.5 Turbo is being touted as OpenAI’s “fastest, most accurate, and most scalable model to date.” The brief description available in the search engine results hints at significant improvements over previous versions, including a substantial increase in the model’s context window to 256,000 tokens, effectively doubling the capacity of the GPT-4 Turbo’s 128,000 tokens. This enhancement, equating to around 200,000 words, could markedly improve the model’s utility for analyzing large documents by allowing it to process and retain more information simultaneously, even over the new and improved models unveiled in January.

The bigger question is about how those tokens will be processed. Existing models with large context windows have been noted for their tendency to overlook portions of the input data. If GPT-4.5 Turbo can address that issue, it would be a big win for OpenAI as it seeks to maintain an edge in the market over rapidly developing rivals like Anthropic’s new Claude 3 models and Google’s recently released Gemini 1.5.

Though there’s no specific release date mentioned in the available indexed text, Anthropic might also have motivated a faster release than usual. The text says GPT-4.5 Turbo will be up-to-date with information until June 2024, which might point to a potential release timeframe. OpenAI usually announces and releases models at the same time, but it might be planning to jump in and get some buzz in the wake of Claude 3’s release and purported ability to match or beat GPT-4 in some areas. Plus, March 14 is GPT-4’s anniversary. OpenAI CEO Sam Altman has scheduled an appearance on Lex Fridman’s podcast for the day and teased the arrival of a new AI model that he says will be “worth the wait.”

Follow @voicebotaiFollow @erichschwartz

OpenAI Showcases New Generative AI Models and Lower API Prices, Cures GPT-4 ‘Laziness’

OpenAI Debuts GPT-4 Multi-Modal Generative AI Model With a Sense of Humor

OpenAI Turns ChatGPT into a Voice Assistant That Can See and Understand Images and Speech