OpenAI Dev News: GPT-5.4, Plugins in Codex, and other fun things

Welcome to the OpenAI developer update—your source for what’s shipping, how teams are building, and practical guidance straight from our engineers.

Plan to get out of your IDE and into the real world? Keep an eye on our community page: a Codex Meetup might pop up near you soon

What's new

What shipped, what changed, what to automate next

Build more powerful workflows with GPT-5.4, GPT-5.4 mini, GPT-5.4 nano

Our most intelligent and efficient frontier model yet—built for professional work and long-horizon tasks—plus our most capable small models yet.

GPT-5.4 brings advanced reasoning, strong coding performance, built-in computer use, and a 1M token context window, powering agents that can operate software, analyze large codebases and documents, and execute complex multi-step workflows.

GPT-5.4 mini is better than GPT-5 mini at reasoning, coding, and computer use, while running more than 2× faster. GPT-5.4 nano is ideal for classification, extraction, ranking, and lightweight subagent tasks.

Extend Codex with Plugins

Connect Codex to tools like Slack, GitHub, Linear, Google Drive, and more—including a plugin for Claude Code.

Plugins in Codex bundle apps, connectors, and skills—pairing tool access with standardized instructions so Codex can reliably complete work across systems.

Bring Codex to Windows

Your command center for running and managing agents is now available natively on Windows and Windows Subsystem for Linux (WSL).

The Codex app on Windows includes built-in sandboxing for agent workflows, support for parallel task execution, and a flexible setup across PowerShell and shell environments.

Build kit

Guides, blogs, and things worth cloning

Explore our new library of Codex use casesFrom PR reviews to design-to-code workflows, see how teams are using Codex to automate the busywork and accelerate real engineering workflows.

Write better prompts for coding agentsA practical guide to structuring prompts for Codex models across planning, tool use, and validation—so long-running workflows complete more reliably.

Package repeatable workflows into SkillsSkills in the API turn instructions, scripts, and assets into reusable building blocks that agents can call directly.

Codex + OpenAI APIs in practiceA hands-on session showing how teams are moving beyond pair programming into agentic delegation, where AI systems take on full engineering tasks from planning to execution.

Prompting GPT-5.4 for frontend workWith this guidance and practical techniques, you can get GPT-5.4 to generate distinctive designs that stand out. Building something worth sharing? Submit your app for a chance to be featured in our Showcase gallery.

Insider scoop

Go behind the scenes with OpenAI engineers

From model to agent: Equipping the Responses API with a computer environment

Dive into how agents actually get built, straight from the developers who build them.

“Giving models a computer environment changes the equation—they’re no longer just generating text, they’re actually doing work. Once you can run tools, manage files, and iterate in place, you start to see agents behave less like chatbots and more like systems.” -Bo Xu, find him on X.

OpenAI’s Michael Bolin on Codex, harness engineering, and the real future of coding agents

Hear from our lead for open-source Codex about the engineering layer that makes coding agents actually function. Hint: It’s not just about better models, but also the environment around them.

“Good developers are always looking to optimize their inner loop, but this is a new inner loop that everyone is still figuring out.” -Michael Bolin, find him on threads.

Devs in the wild

From side projects to professional work

Company spotlight: Raindrop

Using the Responses API for background analysis, Raindrop helps teams catch when agents go off the rails in production, before users ever notice. It surfaces unusual behavior, flags failures, and helps developers quickly get to the “why did this happen?” moment.

Explore more stories from developers building with the Responses API over the past year.

Developer spotlight: Derya Unutmaz

Check out the full thread on X from Derya Unatmaz, professor scientist, immunologist, biomedical engineer & biohacker.

Happy building and vibing,

The OpenAI Team

0 comments