GPT-5.3-Codex: OpenAI Unveils a 25% Faster AI Model That Goes Beyond Coding

OpenAI introduced GPT-5.3-Codex on Feb. 5, a new AI model designed to significantly expand what its Codex agent can do on a computer. 

The company says the model is its most capable agentic coding system yet, combining advanced coding skills with stronger reasoning and professional knowledge. According to OpenAI, GPT-5.3-Codex is 25% faster than its predecessor. It’s also built to handle long, complex tasks that can involve research, tool use, and multi-step execution.

“With GPT‑5.3-Codex, Codex goes from an agent that can write and review code to an agent that can do nearly anything developers and professionals can do on a computer,” according to the company.

One of the most striking claims in OpenAI’s release is how the model was developed. The company says GPT-5.3-Codex is the first OpenAI model to have been instrumental in its own creation.

“The Codex team used early versions to debug its own training, manage its own deployment, and diagnose test results and evaluations,” OpenAI wrote, adding that the team was “blown away” by how much it sped up development.

Strong benchmark results

GPT-5.3-Codex achieved top scores on industry benchmarks, including SWE-Bench Pro and Terminal-Bench 2.0. 

The model scored 77.3% on Terminal-Bench 2.0 and 64.7% on OSWorld-Verified, benchmarks that test agents in real computer environment tasks.

OpenAI also emphasized the model’s improved understanding of user intent. When given simple or vague prompts for website creation, the model now defaults to more functional designs with sensible features rather than requiring extensive specification.

The release came just minutes after OpenAI’s rival, Anthropic, announced its own powerful new model, Claude Opus 4.6, underscoring the fierce competition in the AI space.

Expanded capabilities

The model’s reach extends well beyond coding. It can handle debugging, deployment, monitoring, writing product requirement documents, editing copy, user research, conducting tests, analyzing metrics, and creating presentations and spreadsheets.

OpenAI demonstrated strong performance on GDPval, an evaluation the company released in 2025 that measures AI performance on knowledge-work tasks across 44 occupations, including making presentations and spreadsheets.

OpenAI says GPT-5.3-Codex is the first model it classifies as “high capability” for cybersecurity tasks under its Preparedness Framework.

The model has been directly trained to identify software vulnerabilities, and OpenAI says it is deploying its most comprehensive cybersecurity measures to date. These include automated monitoring, trusted access controls, and enforcement pipelines tied to threat intelligence.

Alongside the release, OpenAI announced a Trusted Access for Cyber pilot program and committed $10 million in API credits to support cybersecurity research, particularly for open-source projects and critical infrastructure.

Available now, but the future is ‘Frontier’

GPT-5.3-Codex is now available to paid ChatGPT users in the Codex app, command-line interface, IDE extensions, and on the web. API access is coming soon.

The launch also dovetails with OpenAI’s new enterprise-focused “Frontier” platform, designed to help companies build and manage teams of AI agents. Early adopters include big names like HP, Intuit, and Uber.

For a hands-on look at how leading AI assistants stack up on everyday tasks like emails, planning, and creative prompts, check out this eWeek comparison of Gemini 3 vs ChatGPT 5.2.

The post GPT-5.3-Codex: OpenAI Unveils a 25% Faster AI Model That Goes Beyond Coding appeared first on eWEEK.