OpenAI is believed to be working on an update to its multimodal language model, just over a year after the release of GPT 4. This new version, dubbed GPT-4.5 Turbo, could be launching in the summer of 2024.
A little more appears to be known about this new model, since Android Authority reveals that OpenAI mistakenly published (before deleting) details on what’s potentially to come. In fact, some social media users found and posted the search engine description for an OpenAI blog post that was later removed.
This leak suggests that GPT-4.5 Turbo will, unsurprisingly, be faster, but also more accurate thanks to a much larger knowledge base than its predecessor.
READ: Navigating the AI wave: Promises and pitfalls of ChatGPT in journalism
And that’s the crux of the matter. The more the model is fed from various sources, the more it will be able to respond to different challenges.
Finally, its context window — the amount of text the model is able to process at once — is set to double from 128,000 to 256,000 tokens. Each token is in fact a fragment of text, varying in length but averaging around four to five characters.
The same leak also mentions a “knowledge cutoff” of June 2024, which could mean an official launch this summer.
Amusingly, this slip-up took place on March 14, 2024, exactly one year after the official launch of GPT 4.