AI agents have revolutionized developer workflows, and their next frontier is knowledge work: processing information, solving complex problems, coming up with new ideas and driving innovation.
Codex, OpenAI’s agentic coding application, is enabling this new frontier. It’s now powered by GPT-5.5, OpenAI’s latest frontier model, which runs on NVIDIA GB200 NVL72 rack-scale systems.
Over 10,000 NVIDIANs — across engineering, product, legal, marketing, finance, sales, HR, operations and developer programs — are already using GPT-5.5-powered Codex to achieve, in their words, “mind-blowing” and “life-changing” results.
NVIDIA engineers have had access to GPT-5.5 through the Codex app for a few weeks, and the gains are measurable. Served on GB200 NVL72, which is capable of delivering 35x lower cost per million tokens and 50x higher token output per second per megawatt compared with prior-generation systems — economics that make frontier-model inference viable at enterprise scale.
Debugging cycles that once stretched across days are closing in hours. Experimentation that previously required weeks is turning into overnight progress in complex, multi-file codebases. Teams are shipping end-to-end features from natural-language prompts, with stronger reliability and fewer wasted cycles than earlier models.
OpenAI’s stunning progress is just the latest example of NVIDIA’s work with every frontier model company — not just to accelerate the use of AI agents inside NVIDIA, but to help the company’s partners build the world’s best, lowest cost and most power efficient models for everyone.
As NVIDIA founder and CEO Jensen Huang told employees in a company-wide email urging everyone to use Codex: “Let’s jump to lightspeed. Welcome to the age of AI.”
A Deployment Built for Enterprise Security
Just like humans, every agent needs its own dedicated computer.
To ensure seamless operation within secure enterprise environments, the Codex app supports remote Secure Shell (SSH) connections to approved cloud virtual machines, allowing agents to work with real company data without exposing it externally.
So to ensure maximum security and auditability, NVIDIA IT rolled out cloud virtual machines (VMs) for every employee to run their agent safely. This provides a dedicated sandbox for the agent to operate at its maximum capabilities while maintaining full auditability. Users can control the Codex agent running in the cloud VM from a user interface that every employee is familiar with.
A zero-data retention policy governs NVIDIA’s deployment, and agents access production systems with read-only permissions through command-line interfaces and Skills — the same agentic toolkit NVIDIA uses to run automation workflows across the company.
A Decade of Full-Stack Collaboration
The GPT-5.5 launch and the Codex rollout reflect more than 10 years of collaboration between NVIDIA and OpenAI. The partnership began in 2016, when NVIDIA founder and CEO Jensen Huang hand-delivered the first NVIDIA DGX-1 AI supercomputer to OpenAI’s San Francisco headquarters.
Since then, the two companies have worked closely across the full AI stack.
NVIDIA was a day-zero partner for OpenAI’s gpt-oss open-weight model launch, optimizing model weights for NVIDIA TensorRT-LLM and ecosystem frameworks including vLLM and Ollama.
OpenAI has committed to deploying more than 10 gigawatts of NVIDIA systems for its next-generation AI infrastructure — a buildout that will put millions of NVIDIA GPUs at the foundation of OpenAI’s model training and inference for years ahead.
And OpenAI and NVIDIA are early silicon and co-design partners: OpenAI provides feedback that informs NVIDIA’s hardware roadmap, and in turn gains early access to new architectures. That relationship produced a concrete milestone — the joint bring-up of the first GB200 NVL72 100,000-GPU cluster. The cluster completed multiple large-scale training runs and set a new benchmark for system-level reliability at frontier scale.
GPT-5.5 is the product of that infrastructure running at full strength.
Learn more in OpenAI’s announcement.