{"id":333923,"date":"2025-12-24T21:42:16","date_gmt":"2025-12-24T21:42:16","guid":{"rendered":"https:\/\/www.newsbeep.com\/uk\/333923\/"},"modified":"2025-12-24T21:42:16","modified_gmt":"2025-12-24T21:42:16","slug":"i-tested-the-new-gpt-5-2-it-just-cant-compete-with-google-gemini-3","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/uk\/333923\/","title":{"rendered":"I Tested the New GPT-5.2\u2014It Just Can’t Compete With Google Gemini 3"},"content":{"rendered":"

OpenAI has had a whirlwind second half of 2025, releasing the GPT-5<\/a>, GPT-5.1<\/a>, and now GPT-5.2 models for ChatGPT<\/a>. But Google is more than keeping pace; it most recently launched its 3 Series of Gemini<\/a> models, which includes Nano Banana Pro<\/a>. Both OpenAI and Google promise more capable, more responsive AI chatbots<\/a> courtesy of these updates. However, based on my testing, you probably won\u2019t notice a difference with GPT-5.2, whereas Gemini 3’s improvements are much more appreciable. Here\u2019s everything you need to know about the latest updates to the biggest chatbots in the world.<\/p>\n

GPT-5.2 Arrived So Quickly Because of Gemini 3<\/p>\n

If it sounds strange that GPT-5.2 arrived less than a month after GPT-5.1, you’re not wrong. OpenAI CEO Sam Altman recently declared a \u201ccode red\u201d<\/a> at the company, ordering employees to focus on immediate improvements to ChatGPT. The first half of Gemini 3 (3 Pro and Nano Banana Pro) came out just days after GPT-5.1, and it beat ChatGPT in various benchmarks. So, I believe OpenAI yanked GPT-5.2 out of the oven before it was fully cooked, just to try and keep up with Gemini.<\/p>\n

Why does Gemini 3 have OpenAI on the back foot? Well, for one, Gemini 3 brought with it Nano Banana Pro, which is the AI image generation model to beat right now. Though it\u2019s not part of the GPT-5.2 package, OpenAI recently launched its own next-generation image model, GPT Image 1.5. I tested this one, and while it’s definitely an improvement over ChatGPT\u2019s previous image generation model, it doesn\u2019t unseat Nano Banana Pro.<\/p>\n

Outside of image generation, Google promises a lot with Gemini 3, which encompasses 3 Flash, 3 Pro, and Nano Banana Pro. It claims that everything from the speed at which Gemini generates responses to the depth and quality of its deep research should be better. Google calls 3 Pro<\/a> its \u201cmost intelligent model\u201d with \u201cfrontier performance across complex reasoning\u201d and says that 3 Flash<\/a> \u201ccombin[es] Gemini 3’s Pro-grade reasoning with Flash-level latency, efficiency and cost.\u201d<\/p>\n

OpenAI similarly promises significant under-the-hood performance upgrades that add up to a more responsive, smarter ChatGPT. Specifically, OpenAI advertises GPT-5.2 as being better at building presentations, completing complex projects, creating spreadsheets, perceiving images, understanding long contexts, and using tools. GPT-5.2\u2019s upgrades are somewhat less wide-ranging than what Google promises with Gemini 3, but they both add up to models that are theoretically more intelligent and responsive.<\/p>\n

Do GPT-5.2 and Gemini 3 Actually Feel Any Different?<\/p>\n

Testing chatbots is a delicate task, as their responses can be inconsistent. They can excel at one thing and struggle with another very similar thing. Furthermore, you might get an excellent response to a prompt and get a worse one if you resend that same prompt. Another complication is that OpenAI’s 5-series of LLMs and Google\u2019s 2.5-series already performed as well as or better than competing models.<\/p>\n

It\u2019s also worth keeping in mind that my impressions with GPT-5.2 and Gemini 3 here are based on my initial hands-on testing. You’ll have to wait for my full, formal reviews of both to get the full picture of their capabilities, as well as more granular comparisons between the two.<\/p>\n

GPT-5.2\u2019s Upgrades Are Subtle, If You Can Even Spot Them<\/p>\n

Keeping all that in mind, I don’t notice much of a difference with GPT-5.2. I\u2019m not a coder or a scientist, but whether I was analyzing images, asking questions about complicated topics, creating presentations or spreadsheets, trying to solve math problems, or using ChatGPT\u2019s Agent<\/a>, the performance and results were very similar to those of GPT-5.1.<\/p>\n

For example, I asked both GPT-5.1 and GPT-5.2 to make me a spreadsheet of the last five Prime Warframes released for the video game Warframe, including the items (relics) necessary to acquire them. GPT-5.1 mistakenly included a Warframe that wasn\u2019t among the five most recent, whereas GPT-5.2 did not. Still, GPT-5.1 quickly fixed the problem after I pointed it out.<\/p>\n

\"GPT-5.1<\/p>\n

\n (Credit: OpenAI)\n<\/p>\n

Overall, however, the response quality of the two models was similar. OpenAI, for example, highlights image recognition improvements with GPT-5.2. But when I asked ChatGPT to identify components in my PC based on a picture I uploaded, both GPT-5.2 and GPT-5.1 got about the same number of things wrong and were similarly vague. This roughly equal performance is what I noticed in most instances, regardless of the particular aspect I was evaluating. I had a similar experience when comparing GPT-5.1 and GPT-5. I often had to refer to the model selector to remind myself which version I was using.<\/p>\n

Gemini 3 Is a Clear Step Forward<\/p>\n

Gemini 3\u2019s upgrades are more obvious. As mentioned, its Nano Banana Pro image model is best-in-class, excelling at both editing and generation. It’s a clear improvement over the original Nano Banana, which earned our Technical Excellence award<\/a>. However, Gemini 3\u2019s upgrades go beyond just what you can do with images.<\/p>\n

For example, 3 Pro performs deep research more quickly and is less likely to waste time going off on unrelated tangents than before. In complex reasoning tests, 2.5 Pro often performed slightly worse than ChatGPT during its 4o and GPT-5 eras, but now, it routinely performs as well as or better than GPT-5.2 across computer science, math, and physics questions. It generally answers more quickly than before, too.<\/p>\n

<\/p>\n

\"Newsletter<\/p>\n

\n Get Our Best Stories!\n <\/p>\n

Your Daily Dose of Our Top Tech News<\/p>\n

\"What's<\/p>\n

Sign up for our What’s New Now newsletter to receive the latest news, best new products, and expert advice from the editors of PCMag.<\/p>\n

Sign up for our What’s New Now newsletter to receive the latest news, best new products, and expert advice from the editors of PCMag.<\/p>\n

\n By clicking Sign Me Up, you confirm you are 16+ and agree to our Terms of Use<\/a> and Privacy
\n Policy<\/a>.\n <\/p>\n

Thanks for signing up!<\/p>\n

Your subscription has been confirmed. Keep an eye on your inbox!<\/p>\n

With 3 Flash, image recognition feels noticeably better. For example, 2.5 Flash struggled to identify as many components with as much detail as ChatGPT when I provided it with a photo of my PC. When I ran this same test with 3 Flash, it provided about the same level of accuracy and detail as GPT-5.2.<\/p>\n

Creative writing is another area of improvement for 3 Flash. When I prompted 2.5 Flash to generate poems, I often felt like they were relatively simplistic, especially compared with ChatGPT’s. 3 Flash\u2019s poems are more interesting and unique, as well as incorporating a wider variety of punctuation.<\/p>\n

GPT-5.2 Shines Mostly for Power Users, But So Can Gemini 3<\/p>\n

To reiterate, I\u2019m not a coder, a scientist, or an executive who makes major decisions for a business. Chances are that most chatbot users don\u2019t fall into these categories either, so Gemini\u2019s enhanced deep research and creative writing abilities are more likely to stand out to more people than GPT-5.2’s performance advances. However, if you do have a more complicated or ultra-specialized use case, you might find GPT-5.2 more exciting.<\/p>\n

CEO of Windsurf Jeff Wang calls GPT-5.2 \u201cthe biggest leap for GPT models in agentic coding since GPT-5\u201d and notes that the \u201cversion bump undersells the jump in intelligence.\u201d CEO of Triple Whale, AJ Orbach, sees \u201cdramatically lower latency\u201d and \u201cmuch stronger tool calling\u201d with GPT-5.2, no longer needing \u201csprawling system prompts because 5.2 will execute cleanly off a simple, one-line prompt.\u201d OpenAI quotes these executives in its blog post<\/a>, so this might well be mere marketing. Still, if you scroll through GPT-5.2 mentions on X<\/a>, you will find many people with niche use cases who are happy with GPT-5.2.<\/p>\n


\n This Tweet is currently unavailable. It might be loading or has been removed.
\n <\/a><\/p>\n

Not to be outdone, Gemini 3 also includes some of the same backend improvements that folks with more specialized interests and needs should appreciate. Google’s series of benchmarks<\/a> for 3 Flash not only show that it’s competitive with GPT-5.2 but also that it outperforms 2.5 Flash to a greater degree than GPT-5.2 compared with GPT-5.1. Meanwhile, OpenAI\u2019s benchmarks of GPT-5.2 Thinking<\/a> show Gemini 3 Pro trailing behind, in many cases, by just a few percentage points and outperforming GPT-5.1 Thinking by the same margin.<\/p>\n

Which Update Is Worth Trying Out?<\/p>\n

If you use either service regularly, there\u2019s no reason not to try out their latest models. In fact, you pretty much have to: Gemini doesn\u2019t allow you to use its legacy models, while ChatGPT allows only paid users to do so. Developers, however, have a more complicated decision to make. GPT-5.2 brings with it a 40% increase in price (per million tokens) over that of GPT-5.1 for API access. Gemini 3\u2019s increase in price is a bit more variable, but it\u2019s still less expensive (and sometimes significantly so) than GPT-5.2.<\/p>\n

AI chatbot model updates don\u2019t have to impact all or even the majority of users to be worthwhile, but GPT-5.2 feels like a lackluster release, especially compared with the tangible response quality improvements with Gemini 3. Although OpenAI\u2019s 5-series of models was already top-notch, GPT-5.2\u2019s improvements are particularly abstract.\u00a0<\/p>\n

Gemini is becoming an increasingly attractive chatbot with each update, while ChatGPT is treading water. But that doesn\u2019t necessarily mean you need to switch to Gemini immediately. The chatbots are quite competitive with each other, after all. Still, Gemini\u2019s updates show more promise.<\/p>\n

Disclosure: Ziff Davis, PCMag’s parent company, filed a lawsuit<\/a> against OpenAI in April 2025, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.<\/p>\n

About Our Expert<\/p>\n

\"Ruben<\/p>\n

Ruben Circelli<\/p>\n

Writer, Software<\/p>\n

Experience<\/p>\n

I\u2019ve been writing about consumer technology and video games for over a decade at a variety of publications, including Destructoid, GamesRadar+, Lifewire, PCGamesN, Trusted Reviews, and What Hi-Fi?, among many others. At PCMag, I review AI and productivity software\u2014everything from chatbots to to-do list apps. In my free time, I\u2019m likely cooking something, playing a game, or tinkering with my computer.<\/p>\n


\n Read Full Bio
\n <\/a><\/p>\n