On This Day, March 14, 2023, OpenAI introduced GPT-4, marking a significant advancement in the field of artificial intelligence. GPT-4, a large multi-modal model capable of understanding both text and image inputs to generate text outputs. It showcased human-level performance on various professional and academic benchmarks. The release represented a leap forward in AI’s capacity to interact, create, and solve problems, pushing the boundaries of what machines can achieve and how they can assist in a wide range of tasks.
A Reddit user discovered a blog post by OpenAI when doing a Google search, which was later removed. OpenAI seems to be working on an upcoming model called GPT-4.5 Turbo, which is expected to surpass the current GPT-4 Turbo in terms of speed, accuracy, and scalability. The leaked information suggests that this new model will have a knowledge cutoff of up to June 2024.
Speculations are arising that OpenAI might release the GPT-4.5 Turbo model in the next three to four months, possibly around July or August, before launching the next-generation GPT-5 model. This development indicates OpenAI’s focus on expanding the context length window, with GPT-4.5 Turbo supporting 256K tokens compared to the current GPT-4 Turbo’s limit of 128K tokens.
