OpenAI has introduced its new GPT-5.2 model, which was trained and deployed on NVIDIA's AI GPUs, including Blackwell & Hopper.
NVIDIA Continues To Speed Up The AI Industry By Using Training & Deploying OpenAI's GPT-5.2 on Its Blackwell & Hopper GPUs
Today, OpenAI introduced its next GPT model, the GPT-5.2, which it calls the most advanced frontier AI model. With this model, OpenAI claims that Enterprise users will save 40-60 minutes a day, and intensive users will be able to save more than 10 hours a week.
GPT‑5.2 Thinking | GPT‑5.1 Thinking | |
| GDPval (wins or ties) Knowledge work tasks | 70.9% | 38.8% (GPT‑5) |
| SWE-Bench Pro (public) Software engineering | 55.6% | 50.8% |
| SWE-bench Verified Software engineering | 80.0% | 76.3% |
| GPQA Diamond (no tools) Science questions | 92.4% | 88.1% |
| CharXiv Reasoning (w/ Python) Scientific figure questions | 88.7% | 80.3% |
| AIME 2025 (no tools) Competition math | 100.0% | 94.0% |
| FrontierMath (Tier 1–3) Advanced mathematics | 40.3% | 31.0% |
| FrontierMath (Tier 4) Advanced mathematics | 14.6% | 12.5% |
| ARC-AGI-1 (Verified) Abstract reasoning | 86.2% | 72.8% |
| ARC-AGI-2 (Verified) Abstract reasoning | 52.9% | 17.6% |
GPT‑5.2 was built in collaboration with our long-standing partners NVIDIA and Microsoft. Azure data centers and NVIDIA GPUs, including H100, H200, and GB200-NVL72, underpin OpenAI’s at-scale training infrastructure, driving significant gains in model intelligence. Together, this collaboration allows us to scale compute with confidence and bring new models to market more quickly.
Powering OpenAI's GPT-5.2 is NVIDIA's AI infrastructure, which includes both its Hopper and Blackwell AI GPU architectures. GPT-5.2 was not only trained but also deployed on these chips & the story does not end there. NVIDIA's Blackwell GPUs are the state-of-the-art AI chips, and they continue to get even better.
With its NVFP4 precision model and the latest optimizations, NVIDIA's Blackwell GPUs are now seeing an impressive 45% gain on the GB200 NVL72 platform in MLPerf v5.1 versus MLPerf v5.0. The Blackwell Ultra platforms sit at 1.9x faster than GB200 NVL72 and up to 4.2x faster than the Hopper H100 solutions. The benchmarks were conducted on Llama 3.1 405B "Training" with a 512-GPU scale.
Not only is NVIDIA achieving faster performance, but also delivering better value. The Blackwell GB200 NVL72 platform is able to offer 90% better training performance per dollar value versus the older H100 solution. And all of that comes with a 3.2x boost in overall training performance.
Compared with the NVIDIA Hopper architecture, NVIDIA GB200 NVL72 systems delivered 3x faster training performance on the largest model tested in the latest MLPerf Training industry benchmarks, and nearly 2x better performance per dollar. And NVIDIA GB300 NVL72 delivers a more than 4x speedup compared with NVIDIA Hopper.
As of right now, NVIDIA's Blackwell and Blackwell Ultra GPUs are widely available across leading cloud service providers, neo-clouds, and server makers. Blackwell Ultra is now rolling out from both server makers and cloud service providers, while Blackwell instances are available through major cloud service providers.
About the author: A Software Engineer by training and a PC enthusiast by passion, Hassan Mujtaba serves as Wccftech's Senior Editor for hardware section. With years of experience in the industry, he specializes in deep-dive technical analysis of next-generation CPU and GPU architectures, motherboards, and cooling solutions. His work involves not only breaking news on upcoming technologies but also extensive hands-on reviews and benchmarking.
Follow Wccftech on Google to get more of our news coverage in your feeds.
