Sam Altman-led OpenAI on 5 February unveiled a brand new Codex mannequin, GPT‑5.3-Codex, which the corporate claims is the “most succesful agentic coding mannequin so far.”
It’s the first mannequin to “meaningfully contribute to its personal creation.” In accordance with the corporate assertion, early variations have been used to debug coaching, handle deployment, and analyse take a look at outcomes and evaluations.
What can the most recent Codex mannequin do?
GPT-5.3-Codex can deal with lengthy, advanced duties that embody analysis, instrument use, and multi-step execution, and it permits coders to work together with the mannequin and information it as it really works, very like collaborating with a colleague, with out shedding context, the corporate assertion added.
Options of GPT-5.3-Codex
OpenAI mentioned that the GPT-5.3-Codex mannequin combines the superior coding talents of the GPT-5.2-Codex with the sturdy reasoning {and professional} data of GPT-5.2 right into a single system. It additional added that the brand new mannequin runs not less than 25% quicker than earlier fashions.
What’s new
With GPT-5.3-Codex, the mannequin has advanced from an agent that used to write down and overview code into one that may carry out almost any process that builders and different professionals do on a pc. Merely put, the Codex mannequin may beforehand write and examine code; nonetheless, with the brand new upgraded model, it might carry out duties just like these of a developer or an expert. This contains researching info, modifying information, managing workflows, and operating instruments.
GPT-5.3-Codex is constructed to assist all work throughout the software program lifecycle, together with debugging, deploying, monitoring, writing Product Necessities Paperwork (PRDs), modifying copy, person analysis, exams, metrics, and extra. It may additionally analyse information in sheets.
Availability
The most recent GPT-5.3-Codex is on the market to all customers with paid ChatGPT plans and can be utilized wherever. This contains the applying, Command Line Interface (CLI), Built-in Improvement Surroundings extension (IDE), and net.
Rival AI mannequin launched by Anthropic
The launch of GPT-5.3-Codex coincided with that of Anthropic’s new AI mannequin, Claude Opus 4.6, because the rivalry between the 2 AI companies heats up. The Opus 4.6 mannequin is an improved model. It’s constructed on its predecessor with stronger coding talents, Anthropic mentioned in an announcement on Thursday. Additional, it fastidiously plans, can deal with long-running agentic duties, works extra reliably in massive codebases, and may provide improved code overview.
OpenAI vs Anthropic
Following the launch of the 2 rival AI fashions, OpenAI’s Sam Altman and Anthropic’s Sholto Douglas, one of many firm’s main researchers appeared on a podcast. In accordance with a report in Enterprise Insiderthe OpenAI chief mentioned, “I feel we will likely be heading in direction of a workflow the place lots of people simply really feel like they’re managing a group of brokers. And because the brokers get higher, they will maintain working at the next and better stage of abstraction.”
Responding to Altman, Douglas mentioned that they’ve in contrast the earlier fashions and have observed key variations. In accordance with him, “The OpenAI fashions had been a bit higher at making an attempt actually, actually, actually onerous on robust issues, however the Anthropic fashions had been a lot quicker and so forth.” He added that whereas OpenAI labored on enhancing pace, Anthropic centered on making the AI fashions “ a lot, significantly better at actually, actually robust issues.”