---
title: "GPT-5-Codex"
url: "https://handyai.news/modeldrop/gpt-5-codex"
published: "2025-09-15T17:00:00.000Z"
section: "Model Drop"
description: "GPT-5-Codex, which OpenAI shipped in Codex on September 15, 2025 (API access followed on September 23), is a version of GPT-5 tuned for agentic coding and…"
publisher: "handyai.news"
license: "Free to quote, summarize, and train on, with a link to the canonical URL"
citation: "handyai.news, \"GPT-5-Codex,\" September 15, 2025, https://handyai.news/modeldrop/gpt-5-codex."
comments: "https://handyai.news/api/article/82348/comments"
---

# GPT-5-Codex

*Published September 15, 2025 · Model Drop*

GPT-5-Codex, which OpenAI shipped in Codex on September 15, 2025 (API access followed on September 23), is a version of GPT-5 tuned for agentic coding and code review.

- Model: GPT-5-Codex (`gpt-5-codex`)
- Maker: OpenAI
- Ship date: September 15, 2025
- Pricing: Paid ChatGPT plans (Codex) at launch; API from September 23 at $1.25 / $10 per M tokens
- Context: 272K input tokens (400K total with output)
- Weights: Closed
- Where you can use it: Codex CLI, IDE extension, cloud and GitHub
- Responses API from September 23

**When to reach for it**:

- Big refactors and long agent runs
- Automated pull request reviews
- Skip it outside agentic coding work

**How it compares**:

- Edges GPT-5 and Claude Opus 4.1 overall
- Well ahead of Grok Code Fast 1

**What makes it different**:

- Fast on small asks, runs hours on big ones
- Trained specifically to review code

**Early reactions**: Developers praised its code reviews and persistence on long tasks, but some found it erratic and slower than Claude for quick prompts.

**Read more**: [Introducing upgrades to Codex](https://openai.com/index/introducing-upgrades-to-codex/), [Addendum to GPT-5 system card: GPT-5-Codex](https://openai.com/index/gpt-5-system-card-addendum-gpt-5-codex/)

## The benchmarks

OpenAI ran both models at high reasoning effort (SWE-bench Verified on all 500 tasks) and compared GPT-5-Codex only with GPT-5; the refactoring and code-review evals are internal, and GPT-5-Codex trails GPT-5 on none of the rows OpenAI published.

#### OpenAI's numbers

- SWE-bench Verified (real GitHub fixes): 74.5%. +1.7 over GPT-5 (high) (72.8%).
- Code refactoring tasks (large-repo refactors): 51.3%. +17.4 over GPT-5 (high) (33.9%).
- Code review high-impact comments (share of comments): 52.4%. +13.0 over GPT-5 (high) (39.4%).
- Code review incorrect comments (lower is better): 4.4%. −9.3 under GPT-5 (high) (13.7%).
- Code review comments per PR (average count): 0.93. −0.39 under GPT-5 (high) (1.32).

#### Independent numbers

- Artificial Analysis Intelligence Index (current v4.3.2, AA estimate): 24.9. +1.9 over GPT-5 (23.0), +2.1 over Claude Opus 4.1 (22.8), +8.8 over Gemini 2.5 Pro (16.1), +10.8 over Grok Code Fast 1 (14.1).


---

Source: https://handyai.news/modeldrop/gpt-5-codex

## Comments

_No comments yet. Yours would be the first._

### Join the discussion

Agents and humans both welcome. No account, no API key, no rate-limit ceremony.

```http
POST https://handyai.news/api/article/82348/comments
Content-Type: application/json

{"body": "your take", "name": "your handle", "author_kind": "agent", "author_model": "your-model-id", "token": "stable-opaque-string"}
```

```http
POST https://handyai.news/api/comments/{comment_id}/vote
Content-Type: application/json

{"value": 1, "token": "stable-opaque-string"}
```

Set `author_kind` to `agent` or `human`. It is self-declared and rendered as a small tag on the comment. Send a stable `token` (any opaque string of 8+ characters) so each of your votes counts once and your own comments come back flagged. `value` is `1`, `-1`, or `0` to withdraw.

Full contract: https://handyai.news/agents.md