OpenAI’s “code pink” response to Google’s Gemini 3 Pro has arrived. On the identical day the corporate introduced a Sora licensing pact with Disney, it took the wraps off GPT-5.2. OpenAI is touting the brand new mannequin as its finest but for real-world, skilled use. “It’s higher at creating spreadsheets, constructing shows, writing code, perceiving photographs, understanding lengthy contexts, utilizing instruments, and dealing with advanced, multi-step initiatives,” stated OpenAI.
In a sequence of 10 benchmarks highlighted by OpenAI, GPT-5.2 Considering, probably the most superior model of the mannequin, outperformed its GPT-5.1 counterpart, generally by a major margin. For instance, in AIME 2025, a check that entails 30 difficult arithmetic issues, the mannequin earned an ideal 100% rating, beating out GPT-5.1’s already state-of-the-art rating of 94 good. It additionally achieved that feat with out turning to instruments like internet search. In the meantime, in ARC-AGI-1, a benchmark that assessments an AI system’s capacity to purpose abstractly like a human being would, the brand new system beat GPT-5.1’s rating by greater than 10 share factors.
OpenAI says GPT-5.2 Considering is best at answering questions factually, with the corporate discovering it produces errors 30 p.c much less steadily. “For professionals, this implies fewer errors when utilizing the mannequin for analysis, writing, evaluation, and choice help — making the mannequin extra reliable for on a regular basis data work,” the corporate stated.
The brand new mannequin ought to be higher in dialog too. Of the model of the system most customers are prone to encounter, OpenAI says “GPT‑5.2 On the spot is a quick, succesful workhorse for on a regular basis work and studying, with clear enhancements in info-seeking questions, how-tos and walk-throughs, technical writing, and translation, constructing on the hotter conversational tone launched in GPT‑5.1 On the spot.“
Whereas it is in all probability overstating issues to counsel it is a make or break launch for OpenAI, it’s honest to say the corporate does have quite a bit driving on GPT 5.2. Its huge launch of 2025, GPT-5, did not meet expectations. Customers complained of a system that generated surprisingly dumb answers and had a boring persona. The frustration with GPT-5 was such that individuals started demanding OpenAI deliver again GPT-4o.
Then got here Gemini 3 Pro — which jumped to the highest of LMArena, a web site the place people price outputs from AI programs to vote on the perfect one. Following Google’s announcement, Sam Altman reportedly referred to as for a “code red” effort to enhance ChatGPT. Earlier than at present, the corporate’s earlier mannequin, GPT-5.1, was ranked sixth on LMArena, with programs from Anthropic and Elon Musk’s xAI occupying the spots between OpenAI between Google.
For an organization that not too long ago signed greater than $1.4 trillion price of infrastructure offers in a bid to outscale the competitors, that was not a great place for OpenAI to be in. In his memo to workers, Altman stated GPT-5.2 can be the equal of Gemini 3 Professional. With the brand new system rolling out now, we’ll see whether or not that is true, and what it’d imply for the corporate if it might probably’t not less than match Google’s finest.
OpenAI is providing three totally different variations of GPT-5.2: On the spot, Considering and Professional. All three fashions will likely be first accessible to customers on the corporate’s paid plans. Notably, the corporate plans to maintain GPT-5.1 round, not less than for a short time. Paid customers can proceed to make use of the older mannequin for the subsequent three months by choosing it from the legacy fashions part.
Trending Merchandise
