OpenAI CEO Sam Altman confirmed recently that GPT-5 will launch in early August. Early reports indicate that GPT-5 will come in three versions: Nano, main, and mini, and that users should be able to access the main and mini versions via OpenAI’s API or ChatGPT, while the nano version will likely be accessed only via the API.
Unlike previous GPT versions that could only accurately answer questions contained within the data used to train the model, without reasoning capabilities, GPT-5 will integrate multiple-step reasoning.
By leveraging technolgies from its large language and reasoning models, OpenAI hopes to achieve an Artificial General Intelligence (AGI) milestone at some point in the future, but that won’t be happening in the short term. Regarding GPT-5, Altman said via X, “We are releasing GPT-5 soon but want to set accurate expectations: this is an experimental model that incorporates new research techniques we will use in future models.” As such, GPT-5 may not perform exactly as expected in every situation and it may be slower to provide responses for complex questions that rely both on the LLM and require reasoning.
we achieved gold medal level performance on the 2025 IMO competition with a general-purpose reasoning system! to emphasize, this is an LLM doing math and not a specific formal math system; it is part of our main push towards general intelligence.
when we first started openai,… https://t.co/X46rspI4l6
— Sam Altman (@sama) July 19, 2025
Altman also announced recently that one of OpenAI’s models acheived gold medal level performance in the IMO (International Mathematical Olympiad) competition, but that model is not GPT-5. Altman doesn’t expect to release a public model that can acheive similar gold medal performance “for many months”.