DeepSeek is the talk of the town right now. Even if you don’t use AI like ChatGPT, which you should totally try, you must have heard all the news about the Chinese startup that tanked the stock market. DeepSeek created AI models that are on par with what ChatGPT can do. The DeepSeek R1 reasoning model is what shocked the world.
R1 can match ChatGPT o1, which didn’t have a rival until now. More importantly, DeepSeek R1 training reportedly costs a fraction of the cost of ChatGPT o1.
The Chinese researchers used software optimization to compensate for the lack of hardware. Certain AI-related components, like high-end NVIDIA components, aren’t available to firms like DeepSeek, at least not officially, due to US sanctions.
According to what OpenAI has said recently, DeepSeek might have also used ChatGPT to train early DeepSeek variants without consent. This could explain the fast and affordable development cycle for the R1 precursors.
But let’s not forget that ChatGPT o3 (or o3-mini) is coming soon. That’s the next-gen reasoning model OpenAI unveiled in December, at the end of its “12 Days” AI event. ChatGPT o3’s release date might be imminent, but it’s not (necessarily) related to the arrival of DeepSeek R1.
Instead, the ChatGPT o3 release should happen all according to schedule. OpenAI said in December that the new reasoning model could be out by the end of January after it’s been tested by third-party researchers.
A couple of weeks ago, Sam Altman took to X to thank external safety researchers for their work on o3-mini testing. The CEO also noted that OpenAI had finalized a version of ChatGPT o3-mini and was beginning the release process.
The exec said the next-gen reasoning model would be released in “~a couple of weeks,” without offering a date. That was on January 17th. We’re now almost two weeks later. I wouldn’t be surprised to see ChatGPT o3-mini come out by the end of this week or, at some point, early next week.
It seems unlikely that both o3 and o3-mini will be released at the same time, though they’re essentially same-generation reasoning models. The latter is cheaper and faster to use. Altman specifically mentioned the o3-mini in his tweets rather than the full o3 model.
Altman also said at the time that o3-mini would be available via the API and ChatGPT. A week later, the exec returned to Twitter with more good news for ChatGPT enthusiasts. He said that the free tier will get access to o3-mini, which is certainy a great development. But ChatGPT Plus users will get a lot more access, with Altman teasing “tons” of usage.
The second ChatGPT o3-mini teaser dropped on January 23rd. OpenAI surprised the world with the Operator AI agent release, and just a day after, the Trump Administration anounced the massive Stargate AI funding project.
Finally, Altman revealed additional details about o3-mini access on Sunday, a day before DeepSeek R1 crashed the market. He said that ChatGPT Plus users will get 100 o3-mini queries per day.
Meanwhile, some people have started teasing the arrival of ChatGPT o3-mini on social media, posting launch date rumors, as seen above. Redditors noticed that an OpenAI employee retweeted the message above, which claims ChatGPT o3-mini will be released on Thursday.
Given all the above, I wouldn’t be surprised to see OpenAI announce ChatGPT o3-mini availability on Thursday or Friday. Worst-case, the next-gen reasoning model would arrive next week, which would still fit OpenAI’s agenda laid out in Deceember and executed since.
Then again, the DeepSeek R1 pressure might be just the kind of motivation OpenAI needs to make o3 available for testing.
