October 25, 2024 – According to a blog post published today by tech media outlet The Verge, OpenAI is planning to unveil a cutting-edge model dubbed “Orion” in December of this year.
Citing the media report, “Orion” will adopt a release strategy distinct from that of GPT-4o and o1, eschewing a widespread launch via ChatGPT. Instead, the model will initially be licensed to close partners, assisting them in building their own products and features.
Furthermore, sources indicate that engineers within Microsoft are preparing to host “Orion” on Azure as early as November.
While Orion is viewed internally at OpenAI as the successor to GPT-4, it remains unclear whether the company will externally refer to it as GPT-5.
OpenAI executives have hinted at the immense power of “Orion,” suggesting that it could be over 100 times more capable than GPT-4. In September of this year, IT Home reported that Tadao Nagasaki, CEO of OpenAI Japan, stated that while GPT-3 and GPT-4 offered incremental performance improvements, GPT Next is expected to deliver a much more significant leap, with a 100-fold increase in performance.
OpenAI’s long-term ambition is to combine its large language models (LLMs), potentially culminating in the creation of a more powerful model known as artificial general intelligence (AGI).
