ChatGPT & OpenAI model release history.

OpenAI has shipped 26 major model releases and 28 other major launches since ChatGPT arrived in November 2022. The latest model recorded here is GPT-6 Sol and GPT-6 Luna, announced 22 Sep 2026. Every entry is dated from OpenAI’s own announcement and links to it.

54 MAJOR RELEASES2022–2026CHECKED 23 SEP 2026
NEWEST GPT FLAGSHIP & SOL & INSTANT, MINI & LUNA
GPT-6 Sol and GPT-6 Luna · 22 Sep 2026
NEWEST IMAGES
ChatGPT Images 2.5 · 8 Sep 2026
NEWEST ASTRA
GPT-6 Astra · 3 Sep 2026
NEWEST OPEN WEIGHTS
gpt-oss open-weight models · 5 Aug 2025
§ 02 — MODELS

Every ChatGPT model release.

26 model releases, newest first. “Current” is the newest release in its line; older ones are superseded, and retired ones have left OpenAI’s apps or API. Where a model is in our benchmark tables, the last column links to its scores, price in AUD and Australian availability.

MODEL RELEASES26 RELEASES
ChatGPT model releases, newest first
RELEASEANNOUNCEDLINESTATUSBENCHMARKS
GPT-6 Sol and GPT-6 LunaGPT flagship & Sol · Instant, mini & LunaCURRENTGPT-6 Sol, GPT-6 Luna
ChatGPT Images 2.5ImagesCURRENT
GPT-6 AstraWATERSHEDAstraCURRENTGPT-6 Astra
GPT-5.6 Sol, Terra and LunaGPT flagship & Sol · Instant, mini & LunaSUPERSEDEDGPT-5.6 Sol, GPT-5.6 Terra, GPT-5.6 Luna
GPT-5.5 InstantInstant, mini & LunaSUPERSEDED
GPT-5.5 in ChatGPTGPT flagship & SolSUPERSEDEDGPT-5.5
ChatGPT Images 2.0ImagesSUPERSEDED
GPT-5.4 miniInstant, mini & LunaSUPERSEDED
GPT-5.4 ThinkingGPT flagship & SolSUPERSEDEDGPT-5.4
GPT-5.3 InstantInstant, mini & LunaSUPERSEDED
ChatGPT ImagesImagesSUPERSEDED
GPT-5.2 in ChatGPTGPT flagship & SolRETIRED
GPT-5.1 in ChatGPTGPT flagship & SolSUPERSEDED
GPT-5 in ChatGPTWATERSHEDGPT flagship & SolSUPERSEDED
gpt-oss open-weight modelsOpen weightsCURRENT
o3 and o4-mini with full tool useo-series reasoningSUPERSEDED
GPT-4.1, 4.1 mini and 4.1 nano in the APIGPT flagship & SolSUPERSEDED
4o image generation in ChatGPTImagesSUPERSEDED
GPT-4.5GPT flagship & SolRETIRED
o3-mini, and reasoning for free userso-series reasoningSUPERSEDED
OpenAI o1-previewWATERSHEDo-series reasoningSUPERSEDED
GPT-4o mini replaces GPT-3.5Instant, mini & LunaSUPERSEDED
GPT-4oWATERSHEDGPT flagship & SolSUPERSEDED
GPT-4 Turbo and the Assistants API at DevDayGPT flagship & SolSUPERSEDED
GPT-4 reaches ChatGPT PlusWATERSHEDGPT flagship & SolRETIRED
ChatGPT research previewWATERSHEDGPT flagship & SolSUPERSEDED
§ 03 — TIMELINE

Every major release, year by year.

54 OF 54

202615 RELEASES

  1. MODEL

    GPT-6 Sol and GPT-6 Luna

    • LAUNCHED
    • CURRENT
    • TIERS AND WORK

    OpenAI added GPT-6 Sol and GPT-6 Luna, cheaper GPT-6 models trained with similar methods to Astra, in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu. Free and Go users can use Luna in the desktop app, and the API models are gpt-6-sol and gpt-6-luna. They are not yet in standard Chat, and API standard pricing is US$2/US$10 per million tokens for Sol and US$0.10/US$0.50 for Luna.

    Why it mattered. API prices are half of GPT-5.6's promotional pricing, which brings GPT-6-generation quality to high-volume and everyday agent work. Astra stays the top tier.

    AVAILABLE TO
    Free · Go · Plus · Pro · Business · Enterprise · Edu · API
  2. PLATFORM

    Agents API (public beta)

    • BETA
    • TIERS AND WORK

    OpenAI released the Agents API in public beta for all developers. It provides the same managed harness and infrastructure that runs Codex, with durable sessions, automatic context compaction and a choice of OpenAI-hosted or third-party sandboxes. There is no extra fee beyond token and tool usage.

    Why it mattered. Developers can run long-running cloud agents on OpenAI's managed Codex harness instead of building their own orchestration, context management and sandboxing.

    AVAILABLE TO
    API
  3. MODEL

    ChatGPT Images 2.5

    • LAUNCHED
    • CURRENT
    • TIERS AND WORK

    OpenAI launched Images 2.5 in ChatGPT, with sharper detail, more precise multi-turn editing and up to 50% lower generation latency than Images 2.0, along with Sketch, templates and on-image comments. In the API it released two models, GPT-Image-2.5 Flare for fast everyday generation and GPT-Image-2.5 Sunburst for precision editing.

    Why it mattered. It upgraded image generation for every ChatGPT tier and gave API developers a choice between a speed-focused model and a precision-focused one.

    AVAILABLE TO
    Free · Go · Plus · Pro · Business · Enterprise · Edu · API
  4. MODEL

    GPT-6 Astra

    • WATERSHED
    • LAUNCHED
    • CURRENT
    • TIERS AND WORK

    OpenAI released GPT-6 Astra, its first GPT-6 generation model, rolling out first to a limited set of organisations and then over the following days to Plus, Pro, Business and Enterprise users, the API, Microsoft Azure and AWS Bedrock. Pro, Business and Enterprise also get GPT-6 Astra Pro (shown as GPT-6 Pro in Chat), and API pricing is US$10 input / US$50 output per million tokens.

    Why it mattered. It is OpenAI's new top model for coding, computer use, research and document work. OpenAI rates it at the Critical threshold for cybersecurity under its Preparedness Framework, so it ships with tighter cyber safeguards and misalignment monitoring that can pause or stop tasks.

    AVAILABLE TO
    Plus · Pro · Business · Enterprise · API
    BENCHMARKS
    GPT-6 Astra →
  5. AGENT

    OpenAI Presence

    • LAUNCHED
    • TIERS AND WORK

    OpenAI launched Presence, an enterprise product for deploying customer-facing and internal voice and chat agents with company-set policies, guardrails, approved actions, escalation rules and a Codex-powered improvement loop. OpenAI deploys it together with each customer.

    Why it mattered. It gave enterprises a packaged OpenAI product for production customer-service and operations agents, instead of building those agents themselves on the API.

    AVAILABLE TO
    Enterprise
  6. MODEL

    GPT-5.6 Sol, Terra and Luna

    • GENERALLY AVAILABLE
    • SUPERSEDED
    • TIERS AND WORK

    After a limited partner preview from 26 June, OpenAI made the GPT-5.6 family generally available in ChatGPT, Codex and the API: Sol (flagship), Terra (balanced) and Luna (lowest cost), plus GPT-5.6 Sol Pro for Pro and Enterprise. Launch API prices per million tokens were $5/$30 for Sol, $2.50/$15 for Terra and $1/$6 for Luna.

    Why it mattered. It brought in the new naming scheme, where the number is the generation and Sol, Terra and Luna are capability tiers that update on their own schedules. It also added the max and ultra (multi-agent) effort settings and was the model behind the launch of ChatGPT Work.

    AVAILABLE TO
    Free · Go · Plus · Pro · Business · Enterprise · API
  7. AGENT

    ChatGPT Work and the new desktop app

    • LAUNCHED
    • TIERS AND WORK

    OpenAI introduced ChatGPT Work, an agent built on Codex technology that researches across connected apps and files and produces finished documents, spreadsheets, presentations, reports and Sites. It also released a new ChatGPT desktop app for macOS and Windows that combines Chat, Work and Codex, and replaced the App Directory with a Plugin Directory.

    Why it mattered. It changed ChatGPT from a chat assistant into an agent that carries out long, multi-step work. On web and mobile it came with paid plans (not Free or Go), and the desktop app gave every plan, including Free, access to Chat, Work and Codex.

    AVAILABLE TO
    Plus · Pro · Business · Enterprise · Edu
  8. MODEL

    GPT-5.5 Instant

    • ROLLING OUT
    • SUPERSEDED
    • PERSONAL DOMAINS

    OpenAI rolls out GPT-5.5 Instant as ChatGPT's new default model for all users, replacing GPT-5.3 Instant.

    Why it mattered. This was the broad consumer-facing GPT-5.5 moment: everyone using default ChatGPT started getting the newer model behaviour.

    AVAILABLE TO
    All ChatGPT users
  9. PRODUCT

    ChatGPT for Excel and Google Sheets

    • LAUNCHED
    • PERSONAL DOMAINS

    OpenAI launches ChatGPT sidebars for Microsoft Excel and Google Sheets so users can build, update, and understand spreadsheets in place.

    Why it mattered. Spreadsheet integrations brought ChatGPT into one of the most important everyday work surfaces for knowledge workers.

    AVAILABLE TO
    Free · Go · Plus · Pro
  10. MODEL

    GPT-5.5 in ChatGPT

    • ROLLING OUT
    • SUPERSEDED
    • PERSONAL DOMAINS

    OpenAI releases GPT-5.5 to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex, with GPT-5.5 Pro for higher-accuracy work.

    Why it mattered. GPT-5.5 is the missing flagship event: it pushed agentic coding, computer use, knowledge work, and research forward inside ChatGPT.

    AVAILABLE TO
    Plus · Pro · Business · Enterprise · Codex
    BENCHMARKS
    GPT-5.5 →
  11. MODEL

    ChatGPT Images 2.0

    • LAUNCHED
    • SUPERSEDED
    • PERSONAL DOMAINS

    OpenAI introduces ChatGPT Images 2.0 across all ChatGPT plans, plus images with thinking for paid users selecting Thinking or Pro models.

    Why it mattered. Images 2.0 made image generation another GPT-5-era model surface and added reasoning time as a creative-control option.

    AVAILABLE TO
    Free · Go · Plus · Pro · Business · Enterprise
  12. MODEL

    GPT-5.4 mini

    • ROLLING OUT
    • SUPERSEDED
    • PERSONAL DOMAINS

    OpenAI rolls out GPT-5.4 mini in ChatGPT as the Thinking feature for Free and Go users and as a fallback when paid users hit GPT-5.4 Thinking limits.

    Why it mattered. Mini model updates matter because they determine the quality of the fallback experience most users see when limits are reached.

    AVAILABLE TO
    Free · Go · Plus · Pro · Enterprise
  13. MODEL

    GPT-5.4 Thinking

    • ROLLING OUT
    • SUPERSEDED
    • PERSONAL DOMAINS

    OpenAI releases GPT-5.4 in ChatGPT as GPT-5.4 Thinking, plus GPT-5.4 Pro, improving professional work, agents, coding, deep research, and computer use.

    Why it mattered. GPT-5.4 was a mainline reasoning jump that folded Codex-style computer-use and professional-work gains back into ChatGPT.

    AVAILABLE TO
    Plus · Team · Pro · Enterprise · Edu · Developers
    BENCHMARKS
    GPT-5.4 →
  14. MODEL

    GPT-5.3 Instant

    • LAUNCHED
    • SUPERSEDED
    • PERSONAL DOMAINS

    OpenAI releases GPT-5.3 Instant, improving answer accuracy, web-search synthesis, refusal judgment, writing range, and conversational flow.

    Why it mattered. GPT-5.3 mattered because it improved the default everyday experience rather than only high-effort reasoning workflows.

    AVAILABLE TO
    ChatGPT users
  15. TRUST & SAFETY

    ChatGPT Health

    • WAITLIST
    • PERSONAL DOMAINS

    OpenAI introduces ChatGPT Health as a dedicated space for health and wellness, with added privacy protections and health data connections.

    Why it mattered. Health showed ChatGPT moving into high-trust personal domains with specialised privacy, data, and evaluation layers.

    AVAILABLE TO
    Free · Go · Plus · Pro

202518 RELEASES

  1. PRODUCT

    ChatGPT Images

    • LAUNCHED
    • SUPERSEDED
    • PERSONAL DOMAINS

    OpenAI introduces an improved ChatGPT Images experience on web and mobile, with a dedicated My images area for generated images.

    Why it mattered. Images moved from isolated generations inside chats to a persistent creative workspace inside ChatGPT.

    AVAILABLE TO
    Free · Go · Plus · Edu · Pro
  2. MODEL

    GPT-5.2 in ChatGPT

    • ROLLING OUT
    • RETIRED
    • PERSONAL DOMAINS

    OpenAI releases GPT-5.2 Instant, Thinking, and Pro in ChatGPT, improving spreadsheets, presentations, code, long context, vision, and tool use.

    Why it mattered. GPT-5.2 was a major capability jump for knowledge work and long-running agentic tasks inside ChatGPT.

    12 Jun 2026: GPT-5.2 Instant, Thinking and Pro were removed from ChatGPT; existing conversations continue on the matching GPT-5.5 models.

    AVAILABLE TO
    Paid · Free · Developers
  3. MODEL

    GPT-5.1 in ChatGPT

    • ROLLING OUT
    • SUPERSEDED
    • PERSONAL DOMAINS

    OpenAI upgrades GPT-5 to GPT-5.1 Instant and GPT-5.1 Thinking, improving conversational style, adaptive reasoning, and user control over tone.

    Why it mattered. GPT-5.1 started the rapid GPT-5.x cadence and made everyday ChatGPT responses feel more natural without abandoning reasoning improvements.

    AVAILABLE TO
    Paid · Free · Enterprise · Edu · Developers
  4. PRODUCT

    ChatGPT Atlas

    • LAUNCHED ON MACOS
    • AGENTIC WORKSPACE

    OpenAI introduces ChatGPT Atlas, a browser designed around ChatGPT as an on-page assistant and action layer.

    Why it mattered. Atlas positioned ChatGPT as a companion across the web, not only a destination at chatgpt.com.

    AVAILABLE TO
    Free · Plus · Pro · Go · Business
  5. PLATFORM

    Apps in ChatGPT

    • PREVIEW
    • AGENTIC WORKSPACE

    OpenAI introduces apps that run inside ChatGPT conversations and releases the Apps SDK preview for developers.

    Why it mattered. This re-opened the platform idea in a richer form: interactive app interfaces inside natural language conversations.

    AVAILABLE TO
    Free · Go · Plus · Pro · Developers
  6. AGENT

    ChatGPT Pulse

    • PREVIEW
    • AGENTIC WORKSPACE

    OpenAI previews Pulse for Pro users on mobile, with personalised daily research cards based on chats, feedback, and connected apps.

    Why it mattered. Pulse moved ChatGPT further into proactive, ambient assistance instead of waiting for a prompt.

    AVAILABLE TO
    Pro mobile
  7. MODEL

    GPT-5 in ChatGPT

    • WATERSHED
    • LAUNCHED
    • SUPERSEDED
    • AGENTIC WORKSPACE

    OpenAI introduces GPT-5 as its smartest, fastest, and most useful model yet, with thinking built into ChatGPT.

    Why it mattered. GPT-5 unified the model story across speed, reasoning, coding, work use cases, and agentic tasks.

    AVAILABLE TO
    Free · Plus · Pro · Business · Enterprise · Developers
  8. MODEL

    gpt-oss open-weight models

    • LAUNCHED
    • CURRENT
    • AGENTIC WORKSPACE

    OpenAI released gpt-oss-120b and gpt-oss-20b, its first open-weight language models since GPT-2, under the Apache 2.0 licence. The larger model runs on a single 80 GB GPU and performs close to o4-mini on core reasoning benchmarks; the smaller one runs on devices with 16 GB of memory.

    Why it mattered. Organisations could download and run an OpenAI reasoning model on their own hardware, which matters for data-residency, privacy and offline use.

    AVAILABLE TO
    Open weights
  9. AGENT

    ChatGPT agent

    • WATERSHED
    • LAUNCHED
    • AGENTIC WORKSPACE

    OpenAI introduces ChatGPT agent, which can reason, browse, use tools, run code, and produce editable deliverables.

    Why it mattered. This combined Operator and deep research into a more general agentic mode inside ChatGPT.

    AVAILABLE TO
    Pro · Plus · Team · Enterprise · Edu
  10. AGENT

    Codex cloud coding agent (research preview)

    • RESEARCH PREVIEW
    • AGENTIC WORKSPACE

    OpenAI launched a research preview of Codex, a cloud-based software engineering agent powered by codex-1 (a version of o3 tuned for coding) that works on many tasks in parallel, each in its own sandbox preloaded with the user's repository. It rolled out first to ChatGPT Pro, Team and Enterprise users, with Plus following on 3 June 2025.

    Why it mattered. It marked OpenAI's move from chat help with code to an agent that writes features, fixes bugs and proposes pull requests on its own.

    AVAILABLE TO
    Pro · Team · Enterprise
  11. MODEL

    o3 and o4-mini with full tool use

    • LAUNCHED
    • SUPERSEDED
    • AGENTIC WORKSPACE

    OpenAI released o3 and o4-mini, the first reasoning models that could use every ChatGPT tool, including web search, Python analysis, image reasoning and image generation, within their thinking. They replaced o1 and o3-mini for Plus, Pro and Team users, with Enterprise and Edu a week later and free users able to try o4-mini via Think.

    Why it mattered. Reasoning models became practical for everyday research and analysis tasks rather than just maths and code puzzles.

    AVAILABLE TO
    Free · Plus · Pro · Team · API
  12. MODEL

    GPT-4.1, 4.1 mini and 4.1 nano in the API

    • LAUNCHED
    • SUPERSEDED
    • AGENTIC WORKSPACE

    OpenAI launched three API-only models, GPT-4.1, GPT-4.1 mini and GPT-4.1 nano, with up to 1 million tokens of context and stronger coding and instruction following than GPT-4o. It also began deprecating GPT-4.5 Preview in the API, with shutdown set for 14 July 2025.

    Why it mattered. Developers got million-token context and cheaper, more reliable models for coding and agents; GPT-4.1 later came to ChatGPT on 14 May 2025.

    AVAILABLE TO
    API
  13. MODEL

    4o image generation in ChatGPT

    • LAUNCHED
    • SUPERSEDED
    • AGENTIC WORKSPACE

    OpenAI built image generation directly into GPT-4o and made it the default image generator in ChatGPT for Free, Plus, Pro and Team users, replacing DALL·E, with Enterprise and Edu to follow. DALL·E remained available through a dedicated GPT, and API access followed on 23 April 2025.

    Why it mattered. Images became far better at following detailed instructions and rendering text, turning ChatGPT into a practical tool for graphics, mock-ups and edits.

    AVAILABLE TO
    Free · Plus · Pro · Team
  14. MODEL

    GPT-4.5

    • RESEARCH PREVIEW
    • RETIRED
    • AGENTIC WORKSPACE

    OpenAI releases GPT-4.5 as a research preview of its strongest GPT model for chat, initially for Pro users and developers.

    Why it mattered. GPT-4.5 showed the continued scaling path for general chat intelligence alongside separate reasoning models.

    26 Jun 2026: Removed from ChatGPT, including custom GPTs (announced 28 May 2026).

    AVAILABLE TO
    Pro · Developers · Plus · Team · Enterprise · Edu
  15. AGENT

    Deep research

    • WATERSHED
    • LAUNCHED
    • AGENTIC WORKSPACE

    OpenAI introduces deep research in ChatGPT for multi-step online research and source-backed reports.

    Why it mattered. Deep research packaged browsing, reasoning, and synthesis into a repeatable workflow for serious information work.

    AVAILABLE TO
    Pro · Plus · Team · Enterprise · Edu · Free
  16. MODEL

    o3-mini, and reasoning for free users

    • LAUNCHED
    • SUPERSEDED
    • AGENTIC WORKSPACE

    OpenAI released o3-mini, a cheaper and faster reasoning model with low, medium and high reasoning-effort settings, in ChatGPT and the API. It replaced o1-mini for paid users, and free users could try it through the Reason button, the first time a reasoning model was offered on the free plan.

    Why it mattered. Step-by-step reasoning for maths, coding and science became available to everyone, including free users, at a much lower cost for developers.

    AVAILABLE TO
    Free · Plus · Team · Pro · API
  17. AGENT

    Operator

    • RESEARCH PREVIEW
    • AGENTIC WORKSPACE

    OpenAI releases Operator, a research preview agent that can use its own browser to click, type, scroll, and complete web tasks.

    Why it mattered. Operator was one of the clearest early signals that ChatGPT would evolve from answering to doing.

    AVAILABLE TO
    Pro US
  18. AGENT

    Scheduled tasks

    • BETA
    • AGENTIC WORKSPACE

    OpenAI begins rolling out scheduled tasks in beta, letting users ask ChatGPT to run one-time or recurring prompts in the future.

    Why it mattered. Tasks were an early step from reactive chat toward proactive assistance.

    AVAILABLE TO
    Plus · Pro · Team

202411 RELEASES

  1. PRODUCT

    Projects in ChatGPT

    • LAUNCHED
    • OMNIMODAL ACCELERATION

    ChatGPT Projects launch for Plus, Team, and Pro users, grouping chats, uploaded files, and custom instructions.

    Why it mattered. Projects made long-running work easier to organise and gave ChatGPT a more durable workspace structure.

    AVAILABLE TO
    Plus · Team · Pro
  2. PLAN

    ChatGPT Pro

    • LAUNCHED
    • OMNIMODAL ACCELERATION

    OpenAI adds ChatGPT Pro, a $200 monthly plan with expanded access to top models and o1 pro mode.

    Why it mattered. Pro formalized compute-intensive frontier usage as a distinct tier for researchers, engineers, and heavy users.

    AVAILABLE TO
    Pro
  3. PRODUCT

    Canvas

    • BETA
    • OMNIMODAL ACCELERATION

    OpenAI introduces canvas, a new interface for writing and coding projects that opens beside the chat.

    Why it mattered. Canvas pushed ChatGPT beyond turn-by-turn conversation into an editable work surface.

    AVAILABLE TO
    Plus · Team · Enterprise · Edu
  4. PRODUCT

    Advanced Voice rolls out to Plus and Team

    • LAUNCHED
    • OMNIMODAL ACCELERATION

    OpenAI began rolling out Advanced Voice, which uses GPT-4o's native audio to hold real-time spoken conversations, to all Plus and Team users in the ChatGPT mobile apps over the course of the week. It was not yet available in the EU, the UK, Switzerland, Iceland, Norway or Liechtenstein, and daily usage limits applied.

    Why it mattered. It made natural, interruptible voice conversation a mainstream ChatGPT feature months after it was first demonstrated with GPT-4o in May 2024.

    AVAILABLE TO
    Plus · Team
  5. MODEL

    OpenAI o1-preview

    • WATERSHED
    • PREVIEW
    • SUPERSEDED
    • OMNIMODAL ACCELERATION

    OpenAI releases the first o1 models in ChatGPT and the API, designed to spend more time reasoning before answering.

    Why it mattered. This split the product into fast general intelligence and deliberate reasoning for harder math, science, and coding work.

    AVAILABLE TO
    Plus · Team · Developers
  6. MODEL

    GPT-4o mini replaces GPT-3.5

    • LAUNCHED
    • SUPERSEDED
    • OMNIMODAL ACCELERATION

    OpenAI released GPT-4o mini, a small, low-cost model priced at 15 US cents per million input tokens and 60 cents per million output tokens, more than 60% cheaper than GPT-3.5 Turbo. In ChatGPT it replaced GPT-3.5 for Free, Plus and Team users from launch, with Enterprise following the next week.

    Why it mattered. Free users got a noticeably better fallback model, and developers got a cheap model with a 128K context window for high-volume tasks.

    AVAILABLE TO
    Free · Plus · Team · Enterprise · API
  7. MODEL

    GPT-4o

    • WATERSHED
    • LAUNCHED
    • SUPERSEDED
    • OMNIMODAL ACCELERATION

    OpenAI introduces GPT-4o and brings more intelligence and tools to free ChatGPT users, including stronger text, vision, and voice capabilities.

    Why it mattered. GPT-4o reset expectations for speed, multimodality, and what the free tier could access.

    AVAILABLE TO
    Free · Plus · Team · Enterprise
  8. PRODUCT

    Memory for Plus users

    • ROLLING OUT
    • PLATFORM TURN

    Memory becomes available to ChatGPT Plus users in supported regions, letting the assistant retain user preferences across conversations.

    Why it mattered. Memory made ChatGPT feel less stateless and opened the path to more personalised assistance.

    AVAILABLE TO
    Plus
  9. PLAN

    ChatGPT Team

    • LAUNCHED
    • PLATFORM TURN

    OpenAI launches ChatGPT Team with a collaborative workspace, admin tools, higher message caps, and business data controls.

    Why it mattered. Team filled the gap between individual Plus accounts and large Enterprise deployments.

    AVAILABLE TO
    Team
  10. PLATFORM

    GPT Store

    • ROLLING OUT
    • PLATFORM TURN

    The GPT Store starts rolling out to Plus, Team, and Enterprise users with popular and featured custom GPTs.

    Why it mattered. It gave the ecosystem a storefront and made public distribution part of the ChatGPT experience.

    AVAILABLE TO
    Plus · Team · Enterprise

20239 RELEASES

  1. PLATFORM

    GPTs

    • WATERSHED
    • LAUNCHED
    • PLATFORM TURN

    OpenAI introduces GPTs: custom versions of ChatGPT with instructions, extra knowledge, and chosen capabilities.

    Why it mattered. GPTs turned customisation into a mainstream product surface and previewed a broader app-like ecosystem.

    AVAILABLE TO
    Plus · Enterprise
  2. MODEL

    GPT-4 Turbo and the Assistants API at DevDay

    • PREVIEW
    • SUPERSEDED
    • PLATFORM TURN

    At its first developer conference OpenAI previewed GPT-4 Turbo, with a 128K-token context window, knowledge to April 2023 and input tokens three times cheaper than GPT-4. It also launched the Assistants API in beta, plus GPT-4 Turbo with vision and the DALL·E 3 API.

    Why it mattered. It made GPT-4-class models far cheaper to run and able to read much longer documents, and gave developers built-in tools for building agent-like assistants.

    AVAILABLE TO
    API
  3. PRODUCT

    Voice and image in ChatGPT

    • ROLLING OUT
    • PLATFORM TURN

    OpenAI begins rolling out voice conversations and image understanding in ChatGPT for Plus and Enterprise users.

    Why it mattered. This changed the interaction model from typed prompts to a multimodal assistant that could understand the world around users.

    AVAILABLE TO
    Plus · Enterprise
  4. PLAN

    ChatGPT Enterprise

    • LAUNCHED
    • LAUNCH GRAVITY

    OpenAI launches ChatGPT Enterprise with business privacy controls, admin features, longer context, and higher-speed GPT-4 access.

    Why it mattered. This marked the move from viral consumer product to sanctioned workplace infrastructure.

    AVAILABLE TO
    Enterprise
  5. PRODUCT

    ChatGPT for iOS

    • LAUNCHED
    • LAUNCH GRAVITY

    OpenAI launches the ChatGPT iOS app with synced conversations, voice input through Whisper, and GPT-4 access for Plus users.

    Why it mattered. Mobile made ChatGPT a daily companion instead of a browser tab people visited only at a desk.

    AVAILABLE TO
    Free · Plus
  6. PLATFORM

    ChatGPT plugins

    • ALPHA
    • LAUNCH GRAVITY

    OpenAI starts rolling out initial plugin support so ChatGPT can access information, run computations, and use third-party services.

    Why it mattered. Plugins introduced the idea that a chat model could become an operating layer across outside services.

    AVAILABLE TO
    Plus · Developers
  7. MODEL

    GPT-4 reaches ChatGPT Plus

    • WATERSHED
    • LAUNCHED
    • RETIRED
    • LAUNCH GRAVITY

    GPT-4 launches with stronger reasoning, reliability, and multimodal research capabilities, with access through ChatGPT Plus.

    Why it mattered. It made ChatGPT feel less like a demo and more like a serious assistant for analysis, coding, and expert work.

    30 Apr 2025: Retired from ChatGPT and replaced by GPT-4o (announced 10 April 2025). GPT-4 stayed available in the API.

    AVAILABLE TO
    Plus · Developers
    SOURCE
    GPT-4 ↗
  8. PLATFORM

    ChatGPT model opens to developers (gpt-3.5-turbo API)

    • LAUNCHED
    • LAUNCH GRAVITY

    OpenAI put the model behind ChatGPT, gpt-3.5-turbo, into its API at US$0.002 per 1,000 tokens, 10 times cheaper than its existing GPT-3.5 models, alongside a Whisper speech-to-text API. It introduced the message-based chat format that became the standard way to call these models.

    Why it mattered. It let any business build ChatGPT-style features into its own apps cheaply, and early adopters included Snapchat, Quizlet, Instacart and Shopify.

    AVAILABLE TO
    API
  9. PLAN

    ChatGPT Plus

    • LAUNCHED
    • LAUNCH GRAVITY

    OpenAI launches ChatGPT Plus with general access during peak times, faster responses, and priority access to new features.

    Why it mattered. Plus created the first consumer business model around ChatGPT and gave OpenAI a release channel for frontier features.

    AVAILABLE TO
    Plus

20221 RELEASE

  1. PRODUCT

    ChatGPT research preview

    • WATERSHED
    • RESEARCH PREVIEW
    • SUPERSEDED
    • LAUNCH GRAVITY

    OpenAI introduces ChatGPT as a conversational research preview built for dialogue, follow-up questions, and feedback.

    Why it mattered. This was the ignition point: a research system became a mass-market interface for general-purpose AI.

    AVAILABLE TO
    Free
§ 04 — ERAS

6 eras of ChatGPT.

  1. ERA 01NOVEMBER 2022AUGUST 2023

    Launch gravity

    ChatGPT goes from research preview to a paid product in nine months: GPT-4, the Plus plan, plugins, the iPhone app and an Enterprise tier.

    7 releases, from ChatGPT research preview
  2. ERA 02SEPTEMBER 2023APRIL 2024

    Platform turn

    Voice and vision, GPTs and the GPT Store, GPT-4 Turbo, the Team plan and memory turn ChatGPT from a chatbot into something people build on and personalise.

    6 releases, from Voice and image in ChatGPT
  3. ERA 03MAY 2024DECEMBER 2024

    Omnimodal acceleration

    GPT-4o, o1’s reasoning, Advanced Voice, canvas, search, Projects and the Pro plan reshape ChatGPT into a work surface.

    8 releases, from GPT-4o
  4. ERA 04JANUARY 2025OCTOBER 2025

    Agentic workspace

    Tasks, Operator, deep research, Codex, ChatGPT agent and GPT-5 move ChatGPT from answering questions to doing work, while o3 and gpt-oss widen the model line-up.

    15 releases, from Scheduled tasks
  5. ERA 05NOVEMBER 2025MAY 2026

    Personal domains

    Monthly GPT-5.x upgrades, new image models, Health and ChatGPT for Excel push ChatGPT into specialised parts of people’s lives and work.

    11 releases, from GPT-5.1 in ChatGPT
  6. ERA 06JULY 2026NOW

    Tiers and Work

    Model names split into tiers (Astra, Sol, Terra, Luna), ChatGPT Work and the desktop app turn ChatGPT into an agent for finished work, and GPT-6 arrives.

    7 releases, from ChatGPT Work and the new desktop app
§ 05 — QUESTIONS

ChatGPT releases, answered.

As of our 23 Sep 2026 review, GPT-6 Sol and GPT-6 Luna, announced by OpenAI on 22 Sep 2026. OpenAI added GPT-6 Sol and GPT-6 Luna, cheaper GPT-6 models trained with similar methods to Astra, in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu. Free and Go users can use Luna in the desktop app, and the API models are gpt-6-sol and gpt-6-luna. They are not yet in standard Chat, and API standard pricing is US$2/US$10 per million tokens for Sol and US$0.10/US$0.50 for Luna.

30 Nov 2022: ChatGPT research preview. OpenAI introduces ChatGPT as a conversational research preview built for dialogue, follow-up questions, and feedback.

As of 23 Sep 2026, the newest release in each line is: gpt-oss open-weight models (5 Aug 2025), GPT-6 Astra (3 Sep 2026), ChatGPT Images 2.5 (8 Sep 2026) and GPT-6 Sol and GPT-6 Luna (22 Sep 2026). Older models in those lines have been superseded, and some have been retired from OpenAI’s apps.

54 on this timeline: 26 model releases and 28 product, platform, agent and plan launches. It leaves out incremental updates, settings and feature drops — only releases that changed what ChatGPT could do or who could use it.

GPT-6 Astra: it can be called from AWS Bedrock in Sydney and Microsoft Azure in Australia East, but only through global routing, so requests may be processed outside Australia. GPT-6 Sol: it can be called from Microsoft Azure in Australia East, but only through global routing, so requests may be processed outside Australia. GPT-6 Luna: it can be called from Microsoft Azure in Australia East, but only through global routing, so requests may be processed outside Australia. Our benchmark pages track Australian cloud availability for each model weekly.

Dates are the day OpenAI announced each release, taken from OpenAI’s own site and documentation.

Major releases only: new models, products, plans and capability launches. Retirements are noted on the model they retired.

Checked 23 Sep 2026 · How the current models compare →

MAKE SENSE OF THE NEXT RELEASE

Choose AI improvements around your business needs.