OpenAI Shares Statistics on Codex: Interesting Trends in Usage
OpenAI Shares Statistics on Codex: Interesting Trends in Usage
OpenAI has released new data on Codex usage, highlighting the growing role of agents in user workflows.
openai.com/index/how-agents-are-transforming-work/
Let's take a look:
– Inside OpenAI, Codex now accounts for 99.8% of their output tokens, while ChatGPT only makes up 0.2%.
– Individual users still predominantly use ChatGPT, but in business, tokens in Codex account for approximately 64% of total consumption.
– The top 1% of the most active Codex users run agents for an average of 71 hours a day (this is the sum across all agents).
– Over time, people are increasingly delegating genuinely long tasks to agents. For instance, prompts for tasks that take about 8 hours of human work now make up more than a quarter of requests. In comparison, in December, they were just over 2%.
Given this, the question arises: are people still vibecoding with agents, or are agents now vibecoding with people?
Why it matters
AnalysisThese statistics indicate that Codex agents are becoming essential tools for businesses, transforming the way long tasks are executed. This could signal a new era of automation in software development.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
AI Radar: GPT-6.1 Terra, Gemini 4 Carbon, Mistral Large 4 and OpenAI/Anthropic pricing changes
A roundup of recent AI news: new LLM releases, OpenAI and Anthropic subscription changes, world models, video generation, and research updates.
Part I. AI Agent Architecture: The Foundations
The eleven foundational building blocks of modern autonomous agents: model taxonomy, the agentic runtime loop, advanced prompting, distillation, skills architecture, memory systems, function calling, and FinOps.
Part II. AI Agent Architecture: From Model to Memory and Tools
A deep engineering dive: tokenization, KV-cache, sampling determinism, the training stack, quantization, context engineering, the safety stack, production RAG, ReAct, and multi-agent swarm systems.
Frontier Models
The most powerful class of artificial intelligence at the forefront of global research (Claude 3.7 Sonnet, OpenAI o3/GPT-4.5, Gemini 2.0 Pro), defining the limits of modern reasoning, autonomy, and coding capabilities.