- Nvidia says OpenAI Codex implementation delivers 50x increase in token production
- Delivers 35x cost reduction compared to GPT-4o
- 10,000 NVIDIA employees now operate codex in human resources, legal, engineering and more
OpenAI has implemented its GPT5.5-powered Codex model at Nvidia in a native enterprise AI implementation.
The model is now available to more than 10,000 Nvidia employees and has reportedly resulted in a 35x cost reduction and a 50x increase in token production per megawatt compared to GPT-4o.
Codex has already achieved “mind-blowing” and “life-changing” results, according to Nvidia employees.
Article continues below.
‘Frontier model inference is viable on an enterprise scale’
The model has been implemented on Nvidia’s GB200 NVL72 rack-scale systems, and this is what has made the huge savings in cost and token production possible, Nvidia claims.
The adoption of Nvidia-Codex goes beyond the typical AI implementations that most companies have seen. Nvidia CEO Jensen Huang said: “Chatbots answer questions. Agents work.”
Codex’s original role was to help GitHub users as a Copilot-style coding assistant, but now it can function as an agent assistant that can perform tasks in non-technical departments such as product, legal, marketing, finance, sales, HR, and operations.
We tried something new with NVIDIA to deploy Codex across an entire company and it was amazing to see it work. Let us know if you would like to do this at your company! pic.twitter.com/Xjn6ShrRuqApril 23, 2026
An email from Jensen Huang, shared by Sam Altman on X, reads: “Incredibly excited everyone [that] You will use Codex to speed up and accomplish work that has never been possible before. Please pass on my congratulations to your team for showing the world the frontier once again. And thank you again for inventing GPT, which gave us the springboard to reason, plan, use tools and more.”
The deployment of Codex on Nvidia systems comes as OpenAI’s rivalry with companies like Anthropic is reaching new heights, especially as Anthropic’s Claude Mythos attracts widespread attention.
While both models have wide-ranging and potentially game-changing applications, Anthropic’s Mythos focuses on patching zero-day vulnerabilities at scale (as demonstrated by Firefox), while OpenAI’s GPT-5.5 Codex can handle real enterprise data in a secure cloud environment, allowing employees to go beyond chatbots and interact with an AI agent capable of securely automating entire workflows.
Follow TechRadar on Google News and add us as a preferred source to receive news, reviews and opinions from our experts in your feeds.
