Capability
n/a
no benchmarks
GitHub stars
74.6k
Harness
ReAct loop
Built-in tools
Environments
1
Terminal
Capability
No public benchmark submissions · yet
Codex CLI has not published benchmark scores on SWE-bench Verified, GAIA, OSWorld or τ-bench. Check the GitHub repo or vendor site for the latest · we track new submissions daily.
Momentum
Awaiting snapshot history · 1 days recorded
GitHub star history appears once daily star tracking has collected enough days.
The Stack
Brain · Harness · Tools · Environment
Brains
Which LLM powers it
gpt-5o3
Harness
The loop pattern
ReAct loop
Tools
How it acts on the world
Built-in
Environments
Where it lives
Terminal
Frequently asked
Pulled from the agents dataset
What is Codex CLI?
Codex is OpenAI's command-line coding agent that runs GPT models locally against your repository with a lightweight REPL-style loop.
What model powers Codex CLI?
Codex CLI runs on gpt-5 and o3. It uses a react loop pattern with built-in tools.
Is Codex CLI open-source?
The harness is open under Apache-2.0 but Codex CLI depends on gpt-5 / o3, which are closed models. You bring your own API key.
How popular is Codex CLI?
Codex CLI has 74,591 stars on GitHub across 404 contributors.
Brains used
GitHub stats
Repoopenai/codex
Stars74,591
Contributors404
Created2025-04-13
Last commit2026-04-11
How we score
Methodology
Composite capability · momentum formula · data sources.
Also in Assistive
Peer agents in the same category · sorted by capability