OpenAI Just Made Codex Way Smarter – Here’s What Changed
OpenAI dropped some big news today: they’ve supercharged their AI coding agent Codex with a custom version of GPT-5. The new GPT-5-Codex can now work on coding tasks for hours instead of minutes, and it’s rolling out to all ChatGPT Plus, Pro, Business, and Enterprise users right now.
I’ve been tracking AI coding tools for months, and this update feels different. Let me break down what’s new and why it matters for your business.
The Big Change: GPT-5-Codex Can Think Longer
The headline feature here is GPT-5-Codex’s “dynamic thinking” ability. Unlike the regular GPT-5 that uses a router to decide upfront how much time to spend on a task, this version can change its mind mid-process.
Alexander Embiricos, OpenAI’s Codex product lead, explained it like this: the model might start working on a coding problem, then five minutes in, decide it needs another hour to solve it properly. In some cases, he’s seen it work for up to seven hours on complex tasks.
That’s a game-changer for serious coding work. Instead of getting a quick, surface-level solution, you get deep, thoughtful code that actually works.
Performance Jumps That Actually Matter
OpenAI shared some benchmarks that caught my attention:
- 90% faster startup times: Codex now caches containers, dropping startup time from 48 seconds to just 5 seconds
- Better code reviews: Software engineers found GPT-5-Codex makes fewer incorrect comments and adds more high-impact suggestions
- Superior refactoring: The model outperforms regular GPT-5 on code refactoring tasks from large repositories
The speed improvement alone is huge. Nobody wants to wait nearly a minute every time they start a coding task.
New Features by Open AI GPT5-Codex That Make Coding Easier
OpenAI packed several practical updates into this release:
Internet Access During Tasks: Codex can now connect to the internet while working on your code. This means it can install dependencies, upgrade packages, and run tests that need external resources. Internet access is off by default, but Pro, Team, and Plus users can enable it with granular control over which domains and methods Codex can access.
Voice Commands: You can now dictate tasks to Codex instead of typing everything. This might sound gimmicky, but I can see it being useful when you’re working through complex problems and want to explain what you need in natural language.
Pull Request Updates: Codex can now update existing pull requests when you follow up on tasks. No more creating multiple PRs for related changes.
Multiple Simultaneous Solutions: Codex can generate several different approaches to the same problem at once, letting you pick the best solution instead of being stuck with the first one it comes up with.
VS Code Integration: There’s a new extension that brings Codex directly into VS Code, Cursor, and other VS Code-based editors. You can sign in with your ChatGPT account without messing around with API keys.
Why GPT5- Codex Matters for Your Business
The AI coding space got crowded fast this year. Cursor hit $500 million in annual recurring revenue, Microsoft’s GitHub Copilot keeps adding features, and Google just launched its own coding agents.
OpenAI clearly felt the pressure to step up their game. This update makes Codex competitive with the best coding tools out there.
For small and medium businesses, this means:
- Faster Development: The 90% speed improvement and dynamic thinking mean your developers spend less time waiting and more time building
- Better Code Quality: The improved code reviews and refactoring capabilities help maintain clean, working codebases
- Lower Learning Curve: Voice commands and better integration make it easier for your team to adopt AI coding tools
- Cost Efficiency: If you’re already paying for ChatGPT Plus ($20/month), you get access to professional-grade coding assistance without additional subscriptions
The Competition Heats Up
This update puts OpenAI back in the conversation with coding-first tools like Cursor and Claude Code. Each platform has its strengths:
- Cursor: Still the speed demon for real-time code completion
- GitHub Copilot: Best integration with existing GitHub workflows
- Claude Code: Strong at understanding complex codebases
- Codex: Now offers the most flexibility with multi-hour problem solving
The real winner here is developers who get to choose from increasingly powerful tools.
What’s Coming Next
OpenAI hinted at deeper integrations coming soon. You’ll be able to assign tasks from the command line, ChatGPT Desktop, or even your issue tracker. They’re also working on mid-task guidance, so you can collaborate with Codex while it’s working instead of just setting it loose.
The company is clearly betting that asynchronous, multi-agent workflows will become the standard way engineers write code. Instead of pair programming with AI in real-time, you’ll delegate longer tasks and check back later.
Should You Try It?
If you’re already using ChatGPT Plus or Pro, there’s no reason not to try the updated Codex. The improvements are substantial, and you’re not paying extra for them.
For teams still on the fence about AI coding tools, this update makes a strong case for giving Codex a shot. The combination of longer thinking time, faster performance, and voice commands addresses many of the complaints I’ve heard about earlier AI coding assistants.
The AI coding wars are far from over, but OpenAI just reminded everyone why they’re still a major player in this space.
Read our most recent articles:
- Sam Altman Just Revealed What GPT-6 Will Do Differently
- OpenAI Brings Open-Weight Models gpt-oss-120b and gpt-oss-20b
- 5 Best AI SEO tools for small businesses in 2025
- Best Free AI Image Generators Without Restrictions in 2025
- Is Character AI safe for kids in 2025?
- Top 5 best characters to talk to on Character.Ai (2025)
Get the latest news and updates from the world of Artificial Intelligence with our weekly newsletter, Artificial Tracker.
