GPT-5 Codex
GPT-5 Codex by OpenAI: a GPT 5 variant optimized for agentic coding, long-horizon technical work, large context windows, and integrated IDE/CLI workflows.
Model Overview
Capabilities, design details, and architectural traits
GPT-5 Codex - Codex-native agentic coding model
GPT-5 Codex is a documented variant of GPT 5 that is explicitly optimized to run as an agent inside Codex products (CLI, IDE extension, web, mobile) and to support practical software-engineering workflows. It is intended for long-horizon, agentic coding tasks rather than generic chat-first use.
| Trait | What the documentation says that makes it distinctive |
|---|---|
| Codex-native | Described as a version of GPT 5 optimized for agentic coding within Codex products rather than the general-purpose GPT 5 model. |
| Agentic long-horizon work | Designed to scale reasoning time dynamically and support extended sessions (from short runs to multi-hour project work) for tasks like refactoring, reviews, and large edits. |
| Integrated IDE/CLI workflows | Provided as the default model across Codex CLI, IDE extensions, web, mobile, GitHub reviews, and terminal workflows to operate directly in developer tools. |
| Large context and output support | Exposed with very large context windows and high max output token limits in the official model spec for handling large repositories and long interactions. |
| Optimized for code review and refactoring | Training and evaluation emphasize debugging, refactors, and code-review quality; documentation cites improved review comment correctness and higher-impact comments. |
Defining concept
The core idea of GPT-5 Codex is a specialized GPT 5 snapshot tuned and delivered as an agent inside developer tooling so it can autonomously navigate repos, run commands, edit files, and persist state across environments for substantial engineering work.
Operational notes
- The model is offered through Codex features in ChatGPT (Plus/Pro/Business/Edu/Enterprise) and via Codex CLI/IDE integrations as the Codex-native option.
- Official materials emphasize its role in improving engineering workflows by being able to choose reasoning time and operate across tools rather than acting as a generic conversational model.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | OpenAI | Anthropic | Anthropic |
| Release Date | September 23, 2025 | September 22, 2026 | September 28, 2026 |
| Knowledge Cutoff | Sep 2024 | - | Jun 2026 |
| Context & Limits | |||
| Context Window | 400K | 1M Best Context Window | 1M Best Context Window |
| Pricing (per 1M tokens) | |||
| Input Pricing | $1.25 Best Input Pricing | $4 | $2 |
| Output Pricing | $10 Best Output Pricing | $20 | $10 Best Output Pricing |
| Modalities | |||
| Inputs | textimage | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 24.9 | 57.6 Best Intelligence Index | 56.0 |
| Coding Index | - | - | - |
| Agentic Index | - | - | - |
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
SciCode Benchmark
Scientific coding and mathematical modeling.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Explore more from OpenAI
Other models by OpenAI
Top AI Models
Leading alternatives by intelligence score
