OpenAI has been quite busy this year, releasing all sorts of exciting AI products, from Operator and its first AI agent to the new 4o image generator model built into ChatGPT and the o3 reasoning chatbot. The latest arrival is the upgraded ChatGPT Agent, which dropped last week, and OpenAI is already experiencing a surge in use. ChatGPT Agent saw massive demand from Pro subscribers, so OpenAI had to postpone the launch for the Plus and Team tiers by a few days.
Despite all that, ChatGPT users are still waiting for OpenAI to release its biggest upgrades of the year. First of all, that means the GPT-5 model that’s supposed to launch this summer. Then there’s also the open-source ChatGPT version that OpenAI has had to postpone twice in the past few months.
GPT-5 is easily the most exciting OpenAI product of the year because it should upgrade the entire chatbot experience. GPT-5 will likely replace GPT-4o, the model that most people use for their interactions with ChatGPT. Interestingly, there’s been an uptick in GPT-5 teasers recently, including a leak that shows GPT-5 is being tested.
Sam Altman’s mysterious coding tool
OpenAI CEO Sam Altman was part of the ChatGPT Agent livestream last week. Then, on Saturday morning, he took to X to tease an unreleased ChatGPT product. “Woke up early on a Saturday to have a couple of hours to try using our new model for a little coding project,” Altman said. “Done in 5 minutes. It is very, very good. Not sure how I feel about it…”
That’s not quite enough to confirm that Altman was using GPT-5 for his coding project. Given that vibe coding is one of the AI capabilities that has seen the most interest over the past couple of years, we expect GPT-5 to feature more advanced coding abilities. It would make sense for Altman to test GPT-5 ahead of the public launch. After all, the CEO said back in spring that GPT-5 would get a summer release.
A few hours before Altman posted the tweet above, well-known X user Tibor Blaho shared a screenshot showing a GPT-5 teaser from Alexander Wei from an X thread where the OpenAI engineer explained how a new OpenAI reasoning model achieved gold medal-level performance at the International Math Olympiad. That teaser is still available on X, reading: “btw, we are releasing GPT-5 soon, and we’re excited for you to try it. But just to be clear: the IMO gold LLM is an experimental research model. We don’t plan to release anything with this level of math capability for several months.”
GPT-5 spotted in testing
Blaho routinely discovers information about unreleased AI models, sharing it on X before the products are released commercially. A few hours after sharing Wei’s teaser, but before Altman dropped his mysterious tweet above, Blaho found evidence of GPT-5 being tested in the wild. An OpenAI model called “openai/gpt-5-reasoning-alpha-2025-07-13” is already in testing. The name for this GPT-5 model indicates it’s a reasoning AI chatbot, similar to ChatGPT o3 and o4-mini that are now available to users.
Given Altman’s previous statements about OpenAI’s plans for ChatGPT, we expect GPT-5 to become the first version where you won’t have to pick the type of experience you want. The AI should be smart enough to automatically switch to reasoning if that’s needed to complete a task. Still, OpenAI might use various models under the GPT-5 umbrella to achieve this performance. They might have different names under the hood.
All of this is speculation, as OpenAI hasn’t provided a clear timeline for the GPT-5 release. Also, delays can always happen, as is the case with the open-source model. More than an hour after Altman’s teaser, OpenAI posted a similar teaser on its X account. “To be clear: We’re releasing GPT-5 soon, but the model we used at IMO is a separate experimental model,” OpenAI said. “It uses new research techniques that will show up in future models—but we don’t plan to release a model with this level of capability for many months.”
