OpenAI launched GPT-5.2 in December and said it was trained and deployed on NVIDIA infrastructure, including NVIDIA Hopper and GB200 NVL72 systems. In February, it released GPT-5.3 Codex, described as the first OpenAI agentic coding model to help build itself, and said it was trained and served entirely on GB200 NVL72.
According to the source text, GPT-5.2 achieved the top reported score for industry benchmarks including GPQA-Diamond, AIME 2025 and Tau2 Telecom. It also set a new state-of-the-art result on ARC-AGI-2, which the text describes as a benchmark targeting skills required to develop AGI.
The source says GPT-5.3 Codex combines the coding performance of GPT‑5.2-Codex and the reasoning capabilities of GPT‑5.2 in one model, with 25% faster performance. It also says the model set new industry highs on SWE-Bench Pro and Terminal-Bench, while showing strong performance on OSWorld and GDPval.
The announcement frames the models as examples of how AI builders train and deploy at scale on NVIDIA’s infrastructure. It says AI model development depends on pretraining, post-training and test-time scaling, and that training frontier models from scratch can require tens of thousands, even hundreds of thousands, of GPUs working together.
The text says NVIDIA GB200 NVL72 delivered 3x faster training performance on the largest model tested in the latest MLPerf Training industry benchmarks, and nearly 2x better performance per dollar than NVIDIA Hopper. It also says NVIDIA GB300 NVL72 delivers a more than 4x speedup compared with NVIDIA Hopper.
Beyond language models, the source says NVIDIA supports AI development across speech, image and video generation, as well as biology and robotics. It cites Evo 2, OpenFold3, Boltz-2 and NVIDIA Clara synthesis models as examples.
The text also says Runway announced Gen-4.5 and GWM-1, both trained on NVIDIA infrastructure and optimized for NVIDIA Blackwell. It adds that Black Forest Labs, Cohere, Mistral, OpenAI, Reflection and Thinking Machines Lab are training on the NVIDIA Blackwell platform.
Finally, the source says NVIDIA Blackwell is widely available from leading cloud service providers, neo-clouds and server makers, and that NVIDIA Blackwell Ultra is now rolling out from server makers and cloud service providers. It names Amazon Web Services, CoreWeave, Google Cloud, Lambda, Microsoft Azure, Nebius, Oracle Cloud Infrastructure and Together AI among providers already offering instances powered by NVIDIA Blackwell.
Source: blogs.nvidia.com.
Companies can share verified announcements through Newz9’s international press release submission page.
