eWEEK content material and product suggestions are editorially impartial. We could generate income whenever you click on on hyperlinks to our companions. Learn More.
OpenAI has launched a brand new AI coding assistant, Codex, constructed to assist builders navigate and automate frequent software program engineering workflows. The device is designed to generate code, repair points, and write checks — all with minimal enter from customers. It operates on the codex-1 mannequin, a variant of OpenAI’s o3 mannequin, personalized particularly for engineering productiveness.
What the Codex Agent Does
Codex features as a built-in companion inside ChatGPT, accessible through a aspect panel. Developers assign duties utilizing easy prompts, selecting between code era or common queries about their codebase. Each request is processed in an remoted, cloud-based surroundings that mirrors the consumer’s undertaking information.
Task length varies relying on scope, with some finishing in beneath a minute and others taking as much as 30 minutes. Once completed, Codex gives a full audit path — capturing what actions it took, outputs from the terminal, and any check outcomes — so builders can evaluate precisely what occurred.
To information the agent’s habits, groups can optionally embody an AGENTS.md file of their undertaking listing. This features as a blueprint for Codex, detailing conventions, testing routines, and structural data. Even within the absence of customized directions, Codex stays efficient, however human oversight stays a key safeguard earlier than integrating its output.
OpenAI additionally emphasised that Codex has been educated to acknowledge and block misuse, significantly makes an attempt to create malicious code. Accompanying its launch, the corporate has up to date its security documentation and analysis standards.
Early Enterprise Adoption
Prior to public rollout, OpenAI granted entry to a handful of corporations for early testing. Cisco, for example, has explored Codex’s potential to streamline software program growth throughout its groups.
Additional early adopters embody:
- Temporal, which has been utilizing the device to create checks and find bugs extra effectively.
- Superhuman, the place product leads can now apply minor code tweaks with out involving builders.
- Kodiak, which is evaluating Codex to speed up enhancements to its autonomous automobile methods.
Who Can Use It, and What It Costs
As of now, Codex is offered to customers on ChatGPT Pro, Team, and Enterprise plans. OpenAI stated entry will quickly increase to customers on Plus and Edu tiers. While use is presently free, the corporate will implement charge limits and introduce usage-based pricing quickly.
For API customers working with the codex-mini-latest mannequin, pricing is structured as follows:
- $1.50 per million enter tokens
- $6.00 per million output tokens
- 75% low cost on repeated prompts (through caching)
Codex stays in preview, and sure limitations apply. It doesn’t but assist image-based duties, and can’t be interrupted mid-task. In some instances, assigning work to Codex could take longer than direct coding. OpenAI stated upcoming variations will enhance interactivity, add mid-task steerage, and combine extra deeply with developer instruments akin to CI methods and challenge trackers.
“Eventually, Codex will feel less like a tool and more like a remote teammate you collaborate with asynchronously,” the corporate stated.
This article was based mostly on reporting by contributor Aminu Abdullahi.
