Anthropic, a tech company, has experienced rapid growth, achieving a valuation of $350 billion in just five years. This success is attributed to Anthropic’s Claude Code, an innovative AI coding tool capable of autonomously generating hundreds of lines of code. However, this advancement has put pressure on competitor OpenAI. In response, OpenAI has enhanced its Codex tool, transforming it from a basic code-completion feature into a sophisticated software engineering ecosystem. Recently, OpenAI introduced its latest GPT-5.5 model on Codex, and it has integrated Codex with NVIDIA’s infrastructure for further experimentation.
In a post, Sam Altman expressed excitement about the successful deployment of Codex across NVIDIA’s operations. The partnership between OpenAI and NVIDIA has yielded tangible benefits, as NVIDIA engineers have been utilizing the GPT-5.5-powered Codex app on NVIDIA GB200 NVL72 systems, resulting in significant improvements in various departments within the company.
Enhanced Performance Through NVIDIA Partnership
NVIDIA confirmed that the integration of GPT-5.5 with Codex on their systems has resulted in significant performance enhancements. The utilization of Codex powered by GPT-5.5 on NVIDIA GB200 NVL72 systems has been praised by over 10,000 NVIDIA employees across various departments for delivering remarkable and transformative outcomes.
Efficiency and Token Advantages
OpenAI’s choice to collaborate with NVIDIA for Codex is speculated to be linked to enhanced token efficiency and positioning itself as a superior alternative to Anthropic. Users of Anthropic have recently experienced faster depletion of their Claude tokens, which serve as a unit of measurement for AI model usage. In contrast, OpenAI’s Codex consumes fewer tokens for similar tasks, making it more efficient and capable. NVIDIA highlighted that their system running Codex offers significantly lower costs per million tokens and higher token output per second per megawatt compared to previous systems.
NVIDIA emphasized the economic benefits of their system, highlighting the substantial cost savings and increased token output efficiency. They noted that tasks that previously took days to complete are now resolved within hours, showcasing the accelerated progress facilitated by Codex.
Industry Support and Future Prospects
NVIDIA’s CEO, Jensen Huang