research
[AINews] OpenAI reports median internal Codex output tokens grew 56x in Research, 32x in Customer Support, 27x in Engineering, and 13x in Legal since November 2025.
Developers building AI workflows must plan for rapidly expanding token usage, which impacts costs, latency, and system design—scalable token management is becoming critical.
What happened
According to Latent Space, OpenAI has released internal metrics showing dramatic growth in median Codex output token counts across various departments since November 2025. Research saw a 56x increase, Customer Support 32x, Engineering 27x, and Legal 13x. These figures reflect how deeply Codex has been adopted for generating code and text within the company. The data points to a broader trend: as AI coding assistants mature, organizations are integrating them into more workflows, leading to exponential growth in usage. For builders, this means that the demand for AI-generated output is likely to continue accelerating, requiring robust infrastructure for token management, cost control, and output quality assurance.
Key takeaways
- OpenAI reported median Codex output tokens grew 56x in Research, 32x in Customer Support, 27x in Engineering, and 13x in Legal since November 2025.
- The data was shared by Latent Space, indicating rapid scaling of internal AI code generation.
- Growth was highest in Research, suggesting heavy use for prototyping and analysis.
- Customer Support and Engineering also saw significant increases, reflecting broad adoption.
- Legal's 13x growth, while lower, shows non-technical teams are also leveraging Codex.
Why it matters
Developers building AI workflows must plan for rapidly expanding token usage, which impacts costs, latency, and system design—scalable token management is becoming critical.
This is an original editorial digest by AI Workflow Center. Full reporting at the source:
Read the original on Latent SpaceMore AI news
All news →
Run Your Own AI Directory