OpenAI has released an official research preview for Codex. Read our full match report. This AI coding agent is the most advanced one ever! This powerful new tool helps users do the work of many software engineers at a time. It makes the coding process more efficient and increases productivity. Beginning today, we’re releasing Codex to all ChatGPT Pro, Enterprise, and Team plan subscribers. This new launch represents a big step forward to OpenAI’s deepening arsenal of AI-powered tools.
Codex powers a more specialized model called codex-1. This iterative, more-efficient optimization of OpenAI’s o3 AI reasoning model is finely tuned for a software engineering application. As a result, Codex is able to produce complex coding solutions while achieving high accuracy and efficiency levels. Users can get started with “generous access” to the tool for free. OpenAI does have rate limits it will introduce in the weeks ahead as it tries to control usage and prioritize safety.
One of the most impressive aspects of Codex is its potential for seamless integration with GitHub. This integration allows users to preload their code repositories directly into the Codex environment, making it easier to work on existing projects without starting from scratch. Plus, users can track the status of any tasks they’ve given to Codex, giving them visibility into and control over their coding workflow.
“malicious software” – OpenAI’s blog post
OpenAI’s Product Lead, Alexander Embiricos, stated that safety was paramount in their deployment strategy for Codex. He noted that much of the safety work to develop the O3 model is already compatible with Codex. This will help us to make sure we’re mitigating real risks associated with AI-generated code. This focus on safety is more important than ever, considering the ways in which AI tools can, without careful auditing, unintentionally create the next generation of “malicious software”.
Additionally, Codex simply forbids users from unlocking their computers and using their browsers while Codex runs. This flexibility facilitates a high-efficiency workflow, letting users work in parallel without having to pause. On top of this, o3 model is now the default for the Codex CLI. Therefore, developers are able to access this extremely powerful functionality through OpenAI’s API at very competitive rates ($1.50 per million input tokens, $6 per million output tokens).
Now OpenAI is launching Codex to enhance ChatGPT’s programming skills. This strategic play puts them squarely in the business of developing products aimed directly at developers and software engineers. OpenAI seems to be doubling down on AI as a coding tool. This shift will position the company as a leader in the rapidly developing intersection of artificial intelligence and software development.