OpenAI Launches Codex, a Multitasking AI Coding Agent

OpenAI on Friday rolled out a robust new device for software program builders, as the corporate pushes additional into automating coding duties with AI.

The brand new product, known as Codex, is an AI agent designed to assist programmers write code, repair bugs, and run checks — usually concurrently.

“Technical groups at OpenAI have began utilizing Codex as a part of their each day toolkit,” OpenAI stated in a blogpost. “It’s most frequently utilized by OpenAI engineers to dump repetitive, well-scoped duties, like refactoring, renaming, and writing checks, that may in any other case break focus.”

“When unsure or confronted with check failures, the Codex agent explicitly communicates these points, enabling customers to make knowledgeable choices about methods to proceed,” OpenAI added.

Not like conventional chatbots that reply to prompts and generate responses in largely phrases, AI brokers like Codex can work together with different software program and on-line companies, similar to serving to you with a DoorDash order or reserving a dinner reservation.

The Codex rollout got here after OpenAI launched GPT-4.5 in February. A livestream demo highlighted its improved reasoning, instinct, and lowered hallucinations.

CEO Sam Altman described it as “the primary mannequin that looks like speaking to a considerate particular person,” but in addition stated that its intelligence and nuance comes at steep computational price. As a consequence of GPU shortages, GPT-4.5 was initially obtainable solely to $200-per-month ChatGPT Professional customers.

Codex is now obtainable to subscribers of OpenAI’s ChatGPT Professional plan, which prices $200 a month. OpenAI additionally stated it can ultimately convey Codex to its different premium choices.

OpenAI didn’t instantly reply to a request for feedback.