As expected and following multiple reports, OpenAI officially announced the GPT-5.2 model. For those unaware, it’s a product that the company fast-tracked out of the heat it’s facing from the success of the latest models from Google and Anthropic. Not to forget, the GPT-5.2 is OpenAI’s second biggest update since the company officially launched GPT-5 in August.
GPT-5.2 is here with improved raw performance results
In the announcement, OpenAI describes GPT-5.2 as its “most capable model series yet for professional knowledge work,” and that’s not an exaggeration to be honest. I’m saying that based on the benchmark results the company shared with us. As officially noted, the GPT-5.2 model scores comparatively more when it comes to raw performance, and it is way more improved GPT-5 model. OpenAI’s updated system card shows clear jumps across some popular benchmarks. The most impressive part of this release is how much GPT-5.2 improves on raw performance. :
- GDPval: 70.9%, up from GPT-5’s 38.8%
- SWE-Bench Pro: 55.6%, up from 50.8%
- AIME 2025 math: a perfect 100%, compared to 94%
- FrontierMath (Tier 1–3): 40.3%, up from 31%
Instead of a single model, GPT-5.2 comes as a three-model lineup: GPT-5.2 Instant, GPT-5.2 Thinking, and GPT-5.2 Pro. All three models are designed for different purposes. This includes faster responses to deeper reasoning to full-scale professional workloads. It’s the most segmented release OpenAI has released so far, and it hints at how the company sees AI being used in different contexts going forward.
For the uninitiated, every test listed above taps into different problem-solving skills; be it code repair, mathematical reasoning, or expert-level logic. It’s important to note that OpenAI says that the GPT-5.2 hallucinates less than previous models, making it a handy upgrade for everyday users. So, you can expect the model to feel more trustworthy in long chats or complex tasks.
You may also like to read: Google’s Gemini 3 Sparks Emergency Inside OpenAI — Sam Altman Declares ‘Code Red’
GPT-5.2 should soon show up in ChatGPT for paid subscribers
If you’re a Plus, Pro, Go, Business, or Enterprise user, you may already see GPT-5.2 listed in ChatGPT. If not, don’t worry, you’ll get it eventually. OpenAI has confirmed that GPT-5.2 is rolling out gradually. The company further says this helps keep ChatGPT smooth and reliable during the transition. GPT-5.1 stays around as a legacy model for about three months before OpenAI retires it. If you’re a free user like me, you’ll eventually get limited access to GPT-5.2, but the timing hasn’t been detailed yet. So, you may need to wait a few more weeks. Speaking of models, it’s also worth noting that new OpenAI image models, codenamed “Hazelnut” and “Chestnut,” have been spotted in the wild.
Have you spotted GPT-5.2 in your ChatGPT menu yet? And if you’ve tried it, how does it feel compared to 5.1? Please don’t forget to drop your thoughts in the comments below.
