OpenAI has unveiled its latest innovation, the GPT-5.2 model, set to revolutionize AI performance with training and deployment powered by NVIDIA’s cutting-edge AI GPUs, Blackwell and Hopper. This partnership promises to enhance AI capabilities and deliver unprecedented efficiency gains for users.
NVIDIA’s Breakthrough in AI with GPT-5.2
The new GPT-5.2 model is positioned as the pinnacle of AI technology by OpenAI, promising substantial time savings for both enterprise and intensive users. With this advancement, enterprise users could gain back 40-60 minutes daily, while intensive users might reclaim over 10 hours weekly. This performance leap is backed by significant improvements in various benchmarks, showcasing the model’s prowess in knowledge work, software engineering, and scientific reasoning.
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.
Unleashing AI Power with NVIDIA’s Advanced Infrastructure
NVIDIA’s AI infrastructure plays a crucial role in advancing GPT-5.2, with its Blackwell and Hopper GPU architectures spearheading this development. These state-of-the-art AI chips not only train but also deploy the new model, achieving unprecedented results. NVIDIA’s Blackwell GPUs reflect a remarkable 45% improvement in performance on the GB200 NVL72 platform in MLPerf v5.1 compared to MLPerf v5.0, making them leaders in the field.
The Blackwell Ultra platforms are up to 1.9 times faster than the GB200 NVL72 and reach speeds 4.2 times faster than the Hopper H100 solutions according to benchmarks conducted on Llama 3.1 405B with a 512-GPU scale.
Revolutionary Performance and Value from NVIDIA
NVIDIA is not just pushing the boundaries of speed but also enhancing economic efficiency. The Blackwell GB200 NVL72 platform presents an impressive 90% superior training performance per dollar compared to its predecessor, the H100. This leads to a 3.2x overall training performance boost.
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.
Currently, NVIDIA’s Blackwell and Blackwell Ultra GPUs are broadly accessible through top cloud service providers and server manufacturers. As these innovations roll out, they set a new standard in AI technology, offering unmatched performance and value.