AI Radar

OpenAI models

OpenAI was announced in December 2015 as a non-profit artificial intelligence research company and states its mission as ensuring that artificial general intelligence benefits all of humanity. It develops the GPT language models, the o-series reasoning models and the open-weight gpt-oss models, and offers its models through ChatGPT and the OpenAI API.

Chronology

  1. 2026

    13 models
    1. Released
      GPT Sol · Language, Multimodal, Reasoning GPT-6.1 Sol

      GPT-6.1 Sol is the first model of OpenAI's GPT-6.1 family, a reasoning language model for text and image input released on 29 September 2026 in the API and in ChatGPT Work and Codex for paid and Edu…

      • Rated Critical in cybersecurity, where GPT-6 Sol was rated High, and served with the GPT-6 Astra safeguards stack.
      • Does not support the none reasoning effort level, which GPT-6 Sol supports.
      Primary source Available · 3 news items
    2. Released
      GPT Luna · Language, Multimodal, Reasoning GPT-6 Luna

      GPT-6 Luna is the Luna tier of OpenAI's GPT-6 family, a reasoning language model for text and image input that the API docs describe as meant for focused, high-volume tasks. It was released with…

      • API prices of 0.10 USD input and 0.50 USD output per million tokens, against 0.20 USD and 1.20 USD for GPT-5.6 Luna.
      • Uses the same types of data and training as GPT-6 Astra, with a knowledge cutoff of 18 May 2026 against 16 February 2026 for GPT-5.6 Luna.
      Primary source Available · 2 news items
    3. Released
      GPT Sol · Language, Multimodal, Reasoning GPT-6 Sol

      GPT-6 Sol is the Sol tier of OpenAI's GPT-6 family, a reasoning language model for text and image input that OpenAI describes as a lower-cost alternative to GPT-6 Astra, trained with similar…

      • API prices of 2 USD input and 10 USD output per million tokens, half of the promotional GPT-5.6 Sol price of 4 USD and 20 USD.
      • Uses the same types of data and training as GPT-6 Astra, with a knowledge cutoff of 20 April 2026 against 16 February 2026 for GPT-5.6 Sol.
      Primary source Available · 1 news item
    4. Released
      GPT Astra · Language, Multimodal, Reasoning · Milestone GPT-6 Astra

      GPT-6 Astra is the first model of OpenAI's GPT-6 generation, a reasoning language model for text and image input released on 3 September 2026 in ChatGPT Work, Codex, the API and Amazon Bedrock…

      • Rated Critical in cybersecurity (GPT-5.6 Sol: High); OpenAI added misalignment monitoring to tool-using inference in external deployment.
      • Adds async tool calling, mid-turn steering and changes of reasoning effort mid-conversation in the Responses API, which is also required for tool calling.
      Primary source Available · 19 news items
    5. Released
      GPT Sol · Language, Multimodal, Reasoning GPT-5.6 Sol (August 2026 revision)

      The August 2026 revision of GPT-5.6 Sol is an updated version of that OpenAI reasoning model, released on 6 August 2026 under the same name for ChatGPT conversations only; Codex and ChatGPT Work…

      • In ChatGPT, a slider lets Plus and Pro users choose how much thought the model puts into a response, and one model covers both Instant responses and deeper reasoning.
      • Offered only in ChatGPT conversations, where it replaced GPT-5.5 Instant together with the August GPT-5.6 Luna; Codex and ChatGPT Work kept the July version.
      Primary source Status unknown · no news linked Some details lack sources
    6. Released
      GPT Luna · Language, Multimodal, Reasoning GPT-5.6 Luna

      GPT-5.6 Luna is the smallest tier of OpenAI's GPT-5.6 family, a reasoning language model for text and image input that the API docs describe as meant for cost-sensitive, high-volume work. Released…

      • Context window of 1,050,000 tokens, up from 400,000 in GPT-5.4 nano, the model of the nano tier that Luna corresponds to.
      • Knowledge cutoff of 16 February 2026, against 31 August 2025 for GPT-5.4 nano.
      Primary source Available · 1 news item
    7. Released
      GPT Sol · Language, Multimodal, Reasoning GPT-5.6 Sol

      GPT-5.6 Sol is the flagship of OpenAI's GPT-5.6 family, a reasoning language model for text and image input offered in ChatGPT, Codex, the API and Amazon Bedrock. Previewed with trusted partners on…

      • Adds Programmatic Tool Calling in the API and, in beta, multi-agent orchestration in the Responses API.
      • Adds a max reasoning effort level, persisted reasoning and Pro mode in the API.
      Primary source Available · no news linked
    8. Released
      GPT Terra · Language, Multimodal, Reasoning GPT-5.6 Terra

      GPT-5.6 Terra is the middle tier of OpenAI's GPT-5.6 family, a reasoning language model for text and image input that OpenAI describes as a capable lower-cost option. Released on 9 July 2026 in…

      • Context window of 1,050,000 tokens, up from 400,000 in GPT-5.4 mini.
      • Knowledge cutoff of 16 February 2026, against 31 August 2025 for GPT-5.4 mini.
      Primary source Available · no news linked
    9. Released
      GPT Instant · Language, Multimodal GPT-5.5 Instant

      GPT-5.5 Instant is the Instant model of OpenAI's GPT-5.5 generation, released on 5 May 2026 as the default ChatGPT model for all users and offered in the API through the chat-latest alias. OpenAI…

      • OpenAI reports fewer hallucinated and inaccurate claims than GPT-5.3 Instant in its internal evaluations.
      • OpenAI describes shorter, more direct answers with fewer follow-up questions, less formatting and fewer emojis.
      Primary source Status unknown · no news linked
    10. Released
      GPT · Language, Multimodal, Reasoning GPT-5.5

      GPT-5.5 is an OpenAI reasoning model for text and image input, released on 23 April 2026 in ChatGPT as GPT-5.5 Thinking and in Codex, with the API a day later. OpenAI describes it as a model for…

      • OpenAI says it matches GPT-5.4's per-token latency in real-world serving while using fewer tokens for the same Codex tasks.
      • Knowledge cutoff moves to 1 December 2025, from 31 August 2025 for GPT-5.4.
      Primary source Available · no news linked
    11. Released
      GPT mini · Language, Multimodal, Reasoning GPT-5.4 mini

      GPT-5.4 mini is a small OpenAI model for text and image input, released on 17 March 2026 with GPT-5.4 nano for high-volume workloads, in the API, Codex and ChatGPT. OpenAI says it brings many…

      • OpenAI says it runs more than twice as fast as GPT-5 mini while improving coding, reasoning, multimodal understanding and tool use.
      • Supports tool search and built-in computer use in the API; GPT-5.4 nano supports neither.
      Primary source Available · no news linked
    12. Released
      GPT · Language, Multimodal, Reasoning GPT-5.4

      GPT-5.4 is an OpenAI reasoning model for text and image input, released on 5 March 2026 as GPT-5.4 Thinking in ChatGPT, as gpt-5.4 in the API and in Codex, alongside GPT-5.4 Pro. OpenAI calls it its…

      • Adds native computer use: through the API's built-in computer tool it can issue mouse and keyboard commands in response to screenshots.
      • Context window grows to 1,050,000 tokens in the API, from 400,000 for GPT-5.2.
      Primary source Available · no news linked
    13. Released
      GPT Instant · Language, Multimodal GPT-5.3 Instant

      GPT-5.3 Instant is an update to the Instant model in OpenAI's ChatGPT for text and image input, released on 3 March 2026 to all ChatGPT users and in the API as gpt-5.3-chat-latest. OpenAI says it…

      • OpenAI says it gives fewer unnecessary refusals and fewer defensive or moralizing caveats than GPT-5.2 Instant.
      • OpenAI says it balances web search results better with its own knowledge, for example to put recent news in context.
      Primary source Retired · no news linked

  2. 2025

    10 models
    1. Released
      GPT · Language, Multimodal, Reasoning GPT-5.2

      GPT-5.2 is an OpenAI model series for text and image input aimed at professional knowledge work, released on 11 December 2025 in ChatGPT as GPT-5.2 Instant, Thinking and Pro and in the API the same…

      • Adds an xhigh reasoning effort level, concise reasoning summaries and context management through compaction.
      • Knowledge cutoff moves to 31 August 2025, from 30 September 2024 for GPT-5.1.
      Primary source Available · 1 news item
    2. Released
      GPT · Language, Multimodal, Reasoning GPT-5.1

      GPT-5.1 is an upgrade of OpenAI's GPT-5 series for text and image input, released in ChatGPT on 12 November 2025 as GPT-5.1 Instant and GPT-5.1 Thinking and in the API a day later. OpenAI says both…

      • OpenAI says thinking time adapts more closely to task difficulty, and GPT-5.1 Instant is the first Instant model that can decide to think before answering.
      • In the API the default reasoning effort is a new setting called none, where GPT-5 defaulted to medium.
      Primary source Deprecated · no news linked
    3. Released
      GPT · Language, Multimodal, Reasoning · Milestone GPT-5

      GPT-5 is a language and reasoning model in OpenAI's GPT line for text and image input, released on 7 August 2025 in ChatGPT and the API. In ChatGPT it is a system of a fast model, a reasoning model…

      • In ChatGPT a router chooses between a fast model and a reasoning model, instead of separate model choices.
      • OpenAI says GPT-5 with thinking needs 50 to 80 percent fewer output tokens than o3 for better results across several capabilities.
      Primary source Deprecated · no news linked
    4. Released
      gpt-oss · Language, Reasoning · Milestone gpt-oss

      gpt-oss is a pair of open-weight language models from OpenAI, gpt-oss-120b and gpt-oss-20b, released on 5 August 2025 under the Apache 2.0 licence with reasoning and tool-use abilities. OpenAI calls…

      Primary source Available · 1 news item
    5. Released
      OpenAI o-series · Language, Multimodal, Reasoning o3-pro

      o3-pro is a version of OpenAI's o3 reasoning model that uses more compute to think longer, released on 10 June 2025 for ChatGPT Pro users and in the API. It accepts text and image input; the docs…

      • Uses more compute than o3 to think longer before answering; the docs say some requests can take several minutes.
      • Limited to ChatGPT Pro users and, according to the docs, the Responses API.
      Primary source Deprecated · no news linked
    6. Released
      OpenAI o-series · Language, Multimodal, Reasoning · Milestone o3

      o3 is a reasoning model in OpenAI's o-series for text and image input, previewed in December 2024 and released with o4-mini on 16 April 2025 in ChatGPT and the API. OpenAI says o3 and o4-mini were…

      • Can use and combine all ChatGPT tools, such as web search and Python, within its reasoning, and decides itself when to call them.
      • Can work with images inside its chain of thought, for example by rotating or zooming them.
      Primary source Deprecated · 1 news item
    7. Released
      OpenAI o-series · Language, Multimodal, Reasoning o4-mini

      o4-mini is a smaller reasoning model in OpenAI's o-series for text and image input, released with o3 on 16 April 2025 in ChatGPT and the API. OpenAI describes it as optimised for fast…

      • Accepts image input; o3-mini accepted text only.
      • Can use and combine ChatGPT tools within its reasoning.
      Primary source Deprecated · no news linked
    8. Released
      GPT · Language, Multimodal GPT-4.1

      GPT-4.1 is a non-reasoning language model in OpenAI's GPT line for text and image input, launched in the API on 14 April 2025 with GPT-4.1 mini and GPT-4.1 nano. OpenAI says the series improves on…

      • Context window of up to one million tokens (1,047,576 in the API docs), larger than GPT-4o's.
      • Maximum output raised to 32,768 tokens, from 16,384 for GPT-4o.
      Primary source Available · 1 news item
    9. Released
      GPT · Language, Multimodal GPT-4.5

      GPT-4.5 is a language model in OpenAI's GPT line for text and image input, released on 27 February 2025 as a research preview in ChatGPT and the API. OpenAI presented it as a step in scaling…

      • Scales pre-training further than GPT-4o, on a mix of public, partner-licensed and in-house data.
      • Uses new alignment techniques that train the larger model with data derived from smaller models; OpenAI says they target steerability and natural conversation.
      Primary source Retired · no news linked
    10. Released
      OpenAI o-series · Language, Reasoning o3-mini

      o3-mini is a small reasoning model in OpenAI's o-series for text input, previewed with o3 in December 2024 and released on 31 January 2025 in ChatGPT and the API. OpenAI calls it its first small…

      • Adds function calling, Structured Outputs and developer messages; OpenAI calls it its first small reasoning model with these features.
      • Adds a choice of low, medium or high reasoning effort; in ChatGPT the high setting is offered as o3-mini-high.
      Primary source Deprecated · no news linked

  3. 2024

    5 models
    1. Released
      OpenAI o-series · Language, Multimodal, Reasoning o1

      o1 is an OpenAI reasoning model, presented on 12 September 2024 and released in ChatGPT on 5 December 2024, when ChatGPT Pro launched with o1 pro mode, a version that uses more compute. A…

      • Accepts image input in the API, where o1-preview took text only.
      • Supports function calling, Structured Outputs and developer messages in the API.
      Primary source Deprecated · no news linked
    2. Released
      OpenAI o-series · Language, Reasoning o1-mini

      o1-mini is a smaller, lower-cost OpenAI reasoning model released on 12 September 2024 together with o1-preview and aimed at STEM tasks, especially math and coding. It was offered to API usage tier 5…

      • Faster than o1-preview and 80 percent cheaper in the API, with higher rate limits in ChatGPT.
      • Narrower knowledge outside STEM, such as dates, biographies and trivia, which OpenAI compares to small models like GPT-4o mini.
      Primary source Retired · no news linked
    3. Released
      OpenAI o-series · Language, Reasoning · Milestone o1-preview

      o1-preview is an OpenAI reasoning model released on 12 September 2024 as an early version of o1 and the first model of the o-series, offered in ChatGPT and to API usage tier 5 developers. OpenAI…

      • Works through a chain of thought before answering, trained with large-scale reinforcement learning; OpenAI says results improve with more time spent thinking.
      Primary source Retired · no news linked
    4. Released
      GPT mini · Language, Multimodal GPT-4o mini

      GPT-4o mini is a small, low-cost OpenAI model released on 18 July 2024 in the API and in ChatGPT, where it took the place of GPT-3.5. It accepts text and image input with a 128K context window, and…

      • Accepts image input as well as text in the API.
      • Context window of 128,000 tokens with up to 16,384 output tokens, up from 16,385 tokens for gpt-3.5-turbo-0125.
      Primary source Available · no news linked
    5. Released
      GPT · Language, Multimodal · Milestone GPT-4o

      GPT-4o is an OpenAI model released on 13 May 2024 in ChatGPT, including the free tier, and in the API. OpenAI describes it as an omni model trained end to end across text, vision and audio; at…

      • OpenAI says it processes audio natively with text and images in one model trained end to end, where the earlier Voice Mode chained three separate models.
      • In the API it is twice as fast and half the price of GPT-4 Turbo, with five times higher rate limits.
      Primary source Available · 1 news item

  4. 2023

    3 models
    1. Released
      GPT · Language, Multimodal GPT-4 Turbo

      GPT-4 Turbo is an OpenAI model for the API, announced at DevDay on 6 November 2023 as a preview of the next generation of GPT-4 and offered that day as gpt-4-1106-preview. It has a 128K context…

      • Context window of 128K tokens, which OpenAI says fits more than 300 pages of text; the gpt-4 API model had 8,192 tokens.
      • Knowledge of world events extends to April 2023.
      Primary source Deprecated · no news linked
    2. Released
      GPT · Language, Multimodal · Milestone GPT-4

      GPT-4 is an OpenAI model that OpenAI describes as a large multimodal model taking image and text input and producing text. It was released on 14 March 2023 in ChatGPT Plus and through an API…

      • Accepts image input in addition to text; at launch image input was tested with a single partner, and it reached ChatGPT in September 2023.
      • OpenAI reports that after six months of alignment work GPT-4 responds less often to requests for disallowed content than GPT-3.5.
      Primary source Deprecated · no news linked
    3. Released
      GPT · Language GPT-3.5 Turbo

      GPT-3.5 Turbo is an OpenAI language model released in the API on 1 March 2023 as gpt-3.5-turbo, which OpenAI said was the model used in the ChatGPT product, and first marketed as the ChatGPT API. It…

      • Makes the ChatGPT model available to developers through the OpenAI API, with dated snapshots such as gpt-3.5-turbo-0301 for pinning a version.
      • OpenAI reported a 90 percent cost reduction for ChatGPT since December 2022 through system-wide optimisations and passed the savings on to API users.
      Primary source Deprecated · no news linked

  5. 2022

    3 models
    1. Released
      GPT · Language · Milestone ChatGPT

      ChatGPT is a conversational language model from OpenAI, fine-tuned from a model in the GPT-3.5 series and offered from 30 November 2022 as a free research preview. OpenAI calls it a sibling model to…

      • Trained for dialogue: human trainers wrote conversations playing both user and assistant, mixed with the InstructGPT dataset converted to dialogue format.
      • Offered as a free research preview on the web instead of through the API.
      Primary source Status unknown · no news linked Some details lack sources
    2. Released
      GPT · Language GPT-3.5

      GPT-3.5 is OpenAI's name for a series of API models trained on a blend of text and code from before the fourth quarter of 2021: the code completion base model code-davinci-002 and the…

      • Trained on a blend of text and code from before the fourth quarter of 2021.
      • text-davinci-002 and code-davinci-002 can insert text into the middle of existing content, not only complete it.
      Primary source Retired · no news linked
    3. Released
      GPT · Language InstructGPT

      InstructGPT is a set of GPT-3 language models that OpenAI fine-tuned with human feedback to follow user instructions, offered through the OpenAI API, where they became the default language models in…

      • Trained with labeler demonstrations, a reward model and PPO to do the task the user asks for, instead of only predicting the next word of internet text.
      • OpenAI says it makes up facts less often than GPT-3 and shows small decreases in toxic output.
      Primary source Retired · no news linked

  6. 2020

    1 model
    1. Released
      GPT · Language · Milestone GPT-3

      GPT-3 is an autoregressive language model from OpenAI with 175 billion parameters, introduced in May 2020 in a paper that applies it to tasks given as text instructions or a few examples in the…

      • 175 billion parameters, against 1.5 billion for the biggest GPT-2 version.
      • Context window doubled to 2048 tokens, from 1024 for GPT-2.
      Primary source Retired · no news linked

  7. 2019

    1 model
    1. Released
      GPT · Language · Milestone GPT-2

      GPT-2 is a transformer language model from OpenAI with up to 1.5 billion parameters, trained on about 40GB of internet text and described by OpenAI as a direct scale-up of GPT. Citing concerns about…

      • More than ten times the parameters of GPT, with 1.5 billion in the biggest version.
      • Trained on WebText, about 40GB of text from 8 million web pages, more than ten times the training data of GPT.
      Primary source Available · no news linked

  8. 2018

    1 model
    1. Released
      GPT · Language · Milestone GPT

      GPT is a transformer language model from OpenAI, published in June 2018 with a paper and released with its code and pretrained weights under the MIT licence. It was pre-trained without supervision…

      Primary source Available · no news linked

Lineage

  1. GPT
    Successors
    GPT-2
  2. GPT-2
    Predecessors
    GPT
    Successors
    GPT-3
  3. GPT-3
    Predecessors
    GPT-2
    Successors
    GPT-3.5
    Variants and derived
    InstructGPT · derived: fine tune
  4. InstructGPT
    Based on
    GPT-3 · derived: fine tune
  5. GPT-3.5
    Predecessors
    GPT-3
    Successors
    GPT-4
    Variants and derived
    ChatGPT · derived: fine tune
  6. ChatGPT
    Successors
    GPT-3.5 Turbo
    Based on
    GPT-3.5 · derived: fine tune
  7. GPT-3.5 Turbo
    Predecessors
    ChatGPT
    Successors
    GPT-4o mini
  8. GPT-4
    Predecessors
    GPT-3.5
    Successors
    GPT-4 Turbo
  9. GPT-4 Turbo
    Predecessors
    GPT-4
    Successors
    GPT-4o
  10. GPT-4o
    Predecessors
    GPT-4 Turbo
    Successors
    GPT-4.1; GPT-4.5; GPT-5
  11. GPT-4o mini
    Predecessors
    GPT-3.5 Turbo
  12. o1-mini
    Successors
    o3-mini
  13. o1-preview
    Successors
    o1
  14. o1
    Predecessors
    o1-preview
    Successors
    o3
  15. o3-mini
    Predecessors
    o1-mini
    Successors
    o4-mini
    Variants and derived
    Phi-4-reasoning (Microsoft) · derived: distillation
  16. GPT-4.5
    Predecessors
    GPT-4o
  17. GPT-4.1
    Predecessors
    GPT-4o
    Successors
    GPT-5
  18. o3
    Predecessors
    o1
    Successors
    GPT-5
    Variants and derived
    o3-pro · variant
  19. o4-mini
    Predecessors
    o3-mini
    Successors
    GPT-5
  20. o3-pro
    Successors
    GPT-5
    Based on
    o3 · variant
  21. gpt-oss

    No recorded relations.

  22. GPT-5
    Predecessors
    GPT-4o; o3; o4-mini; GPT-4.1; o3-pro
  23. GPT-5.1
    Predecessors
    GPT-5
    Successors
    GPT-5.2
  24. GPT-5.2
    Predecessors
    GPT-5.1
  25. GPT-5.3 Instant
    Predecessors
    GPT-5.2
    Successors
    GPT-5.5 Instant
  26. GPT-5.4
    Predecessors
    GPT-5.2
    Successors
    GPT-5.5
  27. GPT-5.4 mini
    Predecessors
    GPT-5
  28. GPT-5.5
    Predecessors
    GPT-5.4
    Successors
    GPT-5.6 Sol
  29. GPT-5.5 Instant
    Predecessors
    GPT-5.3 Instant
  30. GPT-5.6 Luna
    Predecessors
    GPT-5.4 mini
    Successors
    GPT-6 Luna
  31. GPT-5.6 Sol
    Predecessors
    GPT-5.5
    Variants and derived
    GPT-5.6 Sol (August 2026 revision) · revision
  32. GPT-5.6 Terra
    Predecessors
    GPT-5.4 mini
  33. GPT-5.6 Sol (August 2026 revision)
    Based on
    GPT-5.6 Sol · revision
  34. GPT-6 Astra
    Predecessors
    GPT-5.6 Sol
  35. GPT-6 Luna
    Predecessors
    GPT-5.6 Luna
  36. GPT-6 Sol
    Predecessors
    GPT-5.6 Sol
    Successors
    GPT-6.1 Sol
  37. GPT-6.1 Sol
    Predecessors
    GPT-6 Sol

Families

About the organisation

  • Description OpenAI was announced in December 2015 as a non-profit artificial intelligence research company and states its mission as ensuring that artificial general intelligence benefits all of humanity. It develops the GPT language models, the o-series reasoning models and the open-weight gpt-oss models, and offers its models through ChatGPT and the OpenAI API. Primary source[1] [2] [3]

Company · https://openai.com/ · 37 models listed here

Sources

  1. [1] Introducing OpenAI

    Announcement · OpenAI · published

    Provenance
    Primary
    Availability
    Active
    Last checked
    Original ↗ https://openai.com/index/introducing-openai/ Internet Archive capture ↗ captured Founding announcement. openai.com answers scripted requests with a Cloudflare challenge, so the content was read in the Wayback capture of 2026-09-04: dated December 11, 2015, opens by calling OpenAI a non-profit artificial intelligence research company.

  2. [2] Planning for AGI and beyond

    Blog post · OpenAI · published

    Provenance
    Primary
    Availability
    Active
    Last checked
    Original ↗ https://openai.com/index/planning-for-agi-and-beyond/ Internet Archive capture ↗ captured Opens with the mission statement (ensure that artificial general intelligence benefits all of humanity). Read in the Wayback capture of 2026-09-03 because openai.com blocks scripted requests; the page carries an update note dated October 28, 2025.

  3. [3] Models

    Documentation · OpenAI

    Provenance
    Primary
    Availability
    Active
    Last checked
    Original ↗ https://developers.openai.com/api/docs/models API model catalog (HTML title Models | OpenAI API); living page, content as seen on 2026-10-01 via the Markdown version at https://developers.openai.com/api/docs/models.md, which lists GPT models, o-series reasoning models (o1, o3, o4-mini) and the open-weight gpt-oss-120b and gpt-oss-20b. The old address https://platform.openai.com/docs/models redirects here (301).