---
title: "OpenAI launches GPT-5.4 with Thinking and Pro versions"
description: "OpenAI's new model for professional work offers up to 1 million tokens of context and fewer errors than GPT-5.2."
url: https://kurddevelopers.com/en/news/openai-gpt-5-4-pro-thinking
language: en # English
type: news
published: 2026-03-05T18:00:00.000Z
updated: 2026-11-02T00:23:46.000Z
author: "Osamah Ismail"
author_url: https://kurddevelopers.com/en/author/1
section: "AI"
tags: ["OpenAI","GPT-5.4","ChatGPT","AI"]
sources: ["OpenAI","TechCrunch"]
image: https://kurddevelopers.com/uploads/2026/11/5-4-thinking-hero-seo-05a0ba24.png
translations:
  ku: https://kurddevelopers.com/news/openai-gpt-5-4-pro-thinking
  ar: https://kurddevelopers.com/ar/news/openai-gpt-5-4-pro-thinking
publisher: Kurd Developers
publisher_url: https://kurddevelopers.com/
---

# OpenAI launches GPT-5.4 with Thinking and Pro versions

> OpenAI's new model for professional work offers up to 1 million tokens of context and fewer errors than GPT-5.2.

![Image: OpenAI](https://kurddevelopers.com/uploads/2026/11/5-4-thinking-hero-seo-05a0ba24.png)

## What's new?

OpenAI released GPT-5.4 on March 5, calling it "our most capable and efficient frontier model for professional work." It comes in three versions:

- GPT-5.4
- GPT-5.4 Thinking
- GPT-5.4 Pro

## By the numbers

- **Context:** up to 1 million tokens in the API.
- **GDPval:** 83%.
- Record scores on OSWorld-Verified and WebArena Verified.
- Top of Mercor's APEX-Agents leaderboard.

## Fewer errors

According to OpenAI, compared with GPT-5.2:

- 33% less likely to make errors in individual claims.
- 18% fewer responses containing any errors.

## Tool Search

A new Tool Search system lets API models look up tool definitions only when they need them.

**Sources:** OpenAI, TechCrunch

---

Published by [Kurd Developers](https://kurddevelopers.com/) on 2026-03-05. Cite as: “OpenAI launches GPT-5.4 with Thinking and Pro versions”, Kurd Developers, https://kurddevelopers.com/en/news/openai-gpt-5-4-pro-thinking