한국어EN
OpenAI

The GPT family

OpenAI's GPT language models, shown together with ChatGPT, the o-series reasoning models and its open-weight release.

Only versions we have a record for. Missing ones appear as records are added.

23versions2018–2026Latest · GPT-6.1 Sol

Made by AIThe family notes were written by AI (Claude); each version links to its record, where the source is. Published after automated checks; the owner doesn't read every item.How it's made

At a glance

The GPT family — 23 versionsEach GPT version from 2018 to 2026 as a dot, in release order. The list below has the same information.Main lineGPT-1 · Jun 11, 2018GPT-2 · Feb 14, 2019GPT-3 · May 28, 2020InstructGPT · Jan 27, 2022ChatGPT · Nov 30, 2022GPT-4 · Mar 14, 2023GPT-4V(ision) · Sep 25, 2023GPT-4 Turbo · Nov 6, 2023GPT-4o · May 13, 2024GPT-4o mini · Jul 18, 2024GPT-4.5 · Feb 27, 2025GPT-4.1 · Apr 14, 2025GPT-5 · Aug 7, 2025GPT-5.1 · Nov 12, 2025GPT-5.2 · Dec 11, 2025GPT-5.5 · Apr 23, 2026GPT-6 Astra · Sep 3, 2026GPT-6 Sol / GPT-6 Luna · Sep 22, 2026GPT-6.1 Sol · Sep 29, 2026o-series (reasoning)o1-preview / o1-mini · Sep 12, 2024o3-mini · Jan 31, 2025o3 / o4-mini · Apr 16, 2025Open weightsgpt-oss-120b / 20b · Aug 5, 2025
Where labels would overlap, only the dot is drawn — point at it to see the name; every version is in the list below.

23 versions

In release order. Each date comes from the record that documents it — open a name to see the record.

  1. GPT-1

    The first GPT paper: pre-train once, then fine-tune for each task

    Main lineText

  2. GPT-2

    Announced; full model held back over misuse concerns

    Main lineText

  3. GPT-3

    Paper showing tasks solved from examples in the prompt

    Main lineText

  4. InstructGPT

    Announced and made the API's default model

    Main lineText

  5. ChatGPT

    Research preview; chat opened prompting to non-experts

    Main lineText

  6. GPT-4

    Followed instructions better; prompt tricks mattered less

    Main lineTextMultimodal

  7. GPT-4V(ision)

    System card; image understanding moved to wide release

    Main lineMultimodal

  8. GPT-4 Turbo

    Announced with a 128K-token context window

    Main lineText

  9. GPT-4o

    One end-to-end network for text, vision and audio

    Main lineMultimodal

  10. GPT-4o mini

    Released as a low-cost small model

    Main lineMultimodalSmall

  11. o1-preview / o1-mini

    First o-series release: LLMs trained to reason

    o-series (reasoning)Reasoning

  12. o3-mini

    The first reasoning model offered to free ChatGPT users

    o-series (reasoning)ReasoningSmall

  13. GPT-4.5

    Previewed as OpenAI's largest chat model, without a thinking step

    Main lineText

  14. GPT-4.1

    Came with a guide urging explicit, specific prompts

    Main lineText

  15. o3 / o4-mini

    Launched with “thinking with images”

    o-series (reasoning)ReasoningMultimodal

  16. gpt-oss-120b / 20b

    First open-weight language models since GPT-2; Apache 2.0

    Open weightsReasoning

  17. GPT-5

    Launched as a unified, router-based system

    Main lineReasoning

  18. GPT-5.1

    Updated GPT-5 as two models, Instant and Thinking

    Main lineReasoning

  19. GPT-5.2

    Released as Instant, Thinking and Pro

    Main lineReasoning

  20. GPT-5.5

    Paired visual understanding with computer-use scores

    Main lineMultimodal

  21. GPT-6 Astra

    Opened the GPT-6 generation; rolled out in stages

    Main lineReasoning

  22. GPT-6 Sol / GPT-6 Luna

    Two tiers below Astra: mid-range Sol, entry-level Luna

    Main lineReasoning

  23. GPT-6.1 Sol

    Mid-tier update pitched as near-Astra at a lower price

    LatestMain lineReasoning