Today, March 14, marks exactly one year since OpenAI announced GPT-4, its improved language model that brings incremental changes over its predecessor. At launch, the model demonstrated outstanding human-level performance in various professional and academic tests and exams. Earlier today, the company sparked speculation after revealing details about its upcoming GPT-4.5 Turbo large-tongue model on its blog. The paper, which is not currently available, explains that the model outperforms its predecessor (GPT-4) in terms of speed, accuracy, and scalability.
People believe a release in July or August is possible
The most exciting aspect of the leak is the mention of an “information blackout” for the model in June 2024. This means that the AI chatbot will be able to access all information on the internet by next June 2024. While some believe this is a mistake, others believe an OpenAI release is inevitable in July or August before the company moves to the next generation GPT-5.
One of the major updates in GPT-4.5 Turbo is the context length window of 256K tokens, twice the current 128K in GPT-4 Turbo. The move appears to be OpenAI’s response to rivals launching models with larger context windows. Ever since Google renamed Bard to Gemini and introduced new updates to its AI model, the debate has been going on and people are considering considering Gemini as Master. It looks like the new GPT model will solve problems even for OpenAI, even outperforming Google’s AI chatbot.
OpenAI has yet to officially address the leak, and the status of GPT-4.5 Turbo and its June 2024 date remain a mystery. It can’t be a complete coincidence that the leak happened on the first anniversary of GPT-4’s launch (or could it be?). Stay tuned to GizmoChina for future updates on OpenAI.