OpenAI’s GPT releases trace the path to GPT-5.6

What is the OpenAI GPT release timeline, and which model milestones matter most?

This timeline maps OpenAI’s GPT models, products, and platform shifts from GPT-1 to GPT-5.6.

1. The base GPT line

The original GPT lineage is the backbone of OpenAI’s public model history: GPT-1, GPT-2, GPT-3, GPT-3.5, GPT-4, GPT-4 Turbo, GPT-4.1, and the short-lived GPT-4.5 research preview. This is the line that started as text-only research and gradually became the default engine for consumer chat, developer APIs, and enterprise use.

For readers trying to understand the family tree, this lane explains how the company moved from research papers to products people used every day. It also shows the inflection points that mattered most, especially GPT-3’s API debut in 2020 and GPT-4’s multimodal jump in 2023.

• GPT-1: June 2018
• GPT-2: February 2019, staged release
• GPT-3: June 2020, first broad commercial API release
• GPT-4 Turbo: November 2023, 128K context
• GPT-4.1: API-only family in 2025

2. ChatGPT and the consumer breakout

ChatGPT is the milestone that turned GPT from a developer and research story into a mass-market product. Launched on November 30, 2022 as a free research preview built on GPT-3.5, it made conversational AI familiar to a broad audience and set off the adoption wave that followed.

The rest of the consumer timeline shows how OpenAI kept adding surface area around that core experience: ChatGPT Plus, enterprise features, voice, image input, the GPT Store, and Team plans. If you only remember one date, remember the launch of ChatGPT, because it changed the public meaning of GPT.

• 2022-11-30: ChatGPT launch
• 2023-02-01: ChatGPT Plus
• 2023-08-28: ChatGPT Enterprise
• 2023-09-25: voice and image input
• 2024-01-10: GPT Store and ChatGPT Team

3. Codex and the developer tooling shift

Codex has two lives in this timeline. The original 2021 Codex was a GPT-3 descendant tuned for code generation and later retired. The newer Codex, launched in 2025, is a coding agent built on newer model families and tied directly into OpenAI’s current product stack.

That split matters because it shows how OpenAI moved from a single code model to a full coding workflow. The modern Codex story is less about autocomplete alone and more about agentic software tasks, with mainline frontier models shipping directly in Codex by GPT-5.5.

• 2021-08-10: original Codex private beta
• 2021-06-29: GitHub Copilot technical preview
• 2025: Codex returns as a coding agent
• GPT-5-Codex: dedicated coding model line
• Original Codex: retired in 2023

4. The multimodal omni models

GPT-4o marked OpenAI’s cleanest multimodal step before the GPT-5 era. Announced in May 2024, it was designed from the start to handle text, audio, and image in one model, while GPT-4o mini brought that same idea to a smaller, cheaper tier.

This lane is important because it reduced the need to stitch together separate speech, vision, and text systems. GPT-4o also helped normalize real-time interaction, which later fed into the Realtime API and the broader move toward voice-first and agent-style products.

• 2024-05-13: GPT-4o
• 2024-07-18: GPT-4o mini
• 128K-token context window
• Native text, audio, and image handling
• Foundation for Realtime-style interactions

5. The reasoning line and the GPT-5 merge

The o-series introduced a different idea: models that think before answering. Starting with o1-preview and o1-mini in September 2024, OpenAI expanded into o1, o3-mini, o3, and o4-mini, all aimed at deeper reasoning rather than just faster responses.

Then the story converged. On August 7, 2025, GPT-5 unified the fast model and the deeper reasoning model behind a router that decides how much to think. Later releases, especially GPT-5.5 and GPT-5.6, extended that structure with versioned tiers and named capability tiers such as Sol, Terra, and Luna.

• 2024-09-12: o1-preview and o1-mini
• 2025-04: o3 and o4-mini
• 2025-08-07: GPT-5 launch
• 2025-11: GPT-5.1
• 2026-07-09: GPT-5.6 public release

What to pick

If you want the shortest path through the history, start with ChatGPT, GPT-4o, and GPT-5. Those three tell the story of OpenAI’s public shift from chat product to multimodal system to unified reasoning platform.

If you care about product lineage, the base GPT line and Codex matter most. If you care about how OpenAI’s current models are organized, the o-series and GPT-5.6 tiers are the key to reading the modern naming scheme.