OpenAI has begun testing its new o3 and o3-mini models ahead of an expected January 2025 release.
Its o-series models are designed to “reason”, by breaking down prompts into a series of follow-up questions.
The AI tech leader says the o3 and o3-mini will succeed OpenAI’s o1 models and has opened applications for safety testers to improve on the products.
“As models become more capable, we are hopeful that insights from the broader safety community can bring fresh perspectives, deepen our understanding of emerging risks, develop new evaluations, and highlight areas to advance safety research,” OpenAI said.
The o3-mini is slated for release in January, with the o3 itself expected to roll out soon thereafter.
GPT-5 release
OpenAI has also been developing a successor to its flagship ChatGPT model GPT-4.
Not so surprisingly named GPT-5, the new chatbot is expected to be ready for evaluation by key investor Microsoft in mid-2025, yet lower-than-expected results may push that timeline out.
GPT-5 has been in development for more than 18 months and according to the Wall Street Journal, OpenAI has struggled with the software offering.
Its release date is still yet to be confirmed, but OpenAI CEO Sam Altman said last month that it will not roll out before the end of the year.
OpenAI is privately held, and was valued at US$157 billion after a funding round in October.