Codex Scores 75%, Developers Prefer Claude

Rahul Somvanshi

OpenAI launched Codex on May 16, 2025 - a tool that transforms plain English instructions into working computer code, potentially revolutionizing how programmers work.

1. Photo Source: Open AI. Website

Codex operates in its own secure virtual workshop where it can safely write and test code without affecting other systems - taking between 1 to 30 minutes per task.

Photo Source: Edan Meyer. Youtube

Photo Source: ElisaRiva

Tech company Temporal already uses Codex to handle basic coding tasks while their developers focus on solving more complex problems.

1. Photo Source: OpenAI. Youtube

Superhuman discovered an unexpected benefit - their product managers can now make simple code changes without always needing a programmer's help.

1. Photo Source: OpenAI. Youtube

The numbers reveal Codex achieved 75% accuracy on the SWE-Bench Verified benchmark when allowed multiple attempts - helpful but not perfect.

Photo Source: Edan Meyer. Youtube

Carnegie Mellon University doctoral students found that developers often prefer competing tools from companies like Anthropic - is Codex really the best option?

1. Photo Source: Edan Meyer. Youtube

Currently available to ChatGPT Pro, Team, and Enterprise subscribers, Codex will soon expand to other subscription levels with usage limits and pricing.

1. Photo Source: Open AI. Website

Security concerns? Codex works in an isolated environment without internet access - similar to coding in a sealed room.

Photo Source: Nemuel Sereti. Pexels

OpenAI is also upgrading its simpler version, Codex CLI, with faster response times and free API credits for Plus and Pro users.

Photo Source: OpenAI. Youtube

Security concerns? Codex works in an isolated environment without internet access - similar to coding in a sealed room.

The tool can't work with website designs yet and sometimes takes longer than traditional coding - limitations OpenAI acknowledges they're still working on.

Photo Source: Lisa from Pexels

Cisco, one of America's largest tech companies, is already exploring how Codex can help their engineering teams work faster and more efficiently.

1. Photo Source: Open AI. Website

As AI-assisted coding spreads across the tech industry, the key challenge will be balancing automation with human oversight and creativity.

Photo Source: Volker Agueras Gäng (CC BY 2.0)