Skip to content
AI Model Watch

ChatGPT

by OpenAI · openai.com

ChatGPT is OpenAI's AI assistant product, launched November 30, 2022 on GPT-3.5 and powered today by the GPT-5.5 model family — including GPT-5.5 Thinking and Pro variants, the Codex agentic coding agent, the gpt-image-2 image generation model, and ChatGPT Voice. It operates under a freemium model across Free, Plus ($20/mo), Pro ($200/mo), Team, Business, and Enterprise tiers. As of April 2026, ChatGPT counted more than 900 million weekly active users and over 50 million subscribers, making it the most widely used AI assistant globally.

best model GPT-5.5 version 5.5 released Apr 23, 2026

OpenAI was incorporated in December 2015 as a nonprofit AI research organization, co-founded by Sam Altman, Greg Brockman, Ilya Sutskever, Elon Musk, and others, backed by an initial $1 billion pledge. Its earliest published models — GPT-1 (2018), GPT-2 (2019, staged release due to dual-use concerns), and GPT-3 (May 2020, 175 billion parameters) — established the generative pre-trained transformer paradigm. In 2019 Microsoft invested $1 billion and became OpenAI's primary cloud provider; in January 2022 OpenAI published InstructGPT research on aligning language models via human feedback, the direct technical predecessor to ChatGPT.

ChatGPT launched as a free research preview on November 30, 2022, built on GPT-3.5 with a simple chat interface. It reached 100 million monthly active users within two months — the fastest consumer-app adoption in internet history at the time. OpenAI introduced the ChatGPT Plus subscription ($20/month) on February 1, 2023, and followed with GPT-4 on March 14, 2023: a multimodal model accepting text and image input, simultaneously deployed in Microsoft Bing. ChatGPT plugins launched March 23, 2023, offering early tool-use and browsing. Italy banned the service in March 2023 over GDPR and age-verification shortcomings; OpenAI implemented the requested safeguards and the ban was lifted in April 2023.

November 2023 brought OpenAI's most dramatic episode: the board fired Sam Altman on November 17, citing a failure to be 'consistently candid.' Five days later, following intense pressure from Microsoft and a near-total employee revolt, Altman was reinstated and the board was reconstituted. That same month, OpenAI held its first DevDay, unveiling GPT-4 Turbo with a 128K-token context window and the Assistants API. The GPT Store launched in January 2024 with over 3 million user-built GPTs. May 2024 brought GPT-4o ('omni'), folding text, audio, and vision into a single real-time model — and also brought the resignation of Chief Scientist Ilya Sutskever (succeeded by Jakub Pachocki) and safety co-leader Jan Leike, who cited the company's deprioritization of safety. By late 2024, roughly half of OpenAI's AI safety researchers had left the company.

In September 2024, OpenAI released o1-preview, introducing a new 'reasoning' model family that produces internal chain-of-thought before answering. The full o1 model and the $200/month ChatGPT Pro tier launched in December 2024 alongside the Sora text-to-video model. ChatGPT Search went live October 31, 2024, embedding real-time cited web results directly in chat. By August 2024, ChatGPT had 200 million weekly active users. Early 2025 saw the Operator autonomous web agent (January), Deep Research multi-source report synthesis (February), and the Codex cloud-based software engineering agent research preview (May 2025). On August 7, 2025, GPT-5 launched as OpenAI's first 'unified' model — merging the fast GPT-series and the o-series reasoning models behind a real-time router — and became the default for all ChatGPT users, though the rollout was bumpy, requiring OpenAI to restore legacy model access and adjust the model's personality following user complaints. In October 2025, OpenAI completed its conversion to a Public Benefit Corporation (OpenAI Group PBC), with the OpenAI Foundation retaining a 26% governance equity stake.

The GPT-5 generation iterated at a sub-60-day cadence: GPT-5.1 (November 2025), GPT-5.2 (December 2025), GPT-5.4 (March 2026, integrating Codex-trained coding capabilities and native computer use), and GPT-5.5 (April 23, 2026), the current flagship. GPT Image 2 (gpt-image-2), launched April 21, 2026 as ChatGPT Images 2.0, replaced DALL-E 3 and GPT Image 1.5 with a reasoning-powered image model achieving ~99% multilingual text accuracy and up to 4K resolution. As of April 2026, ChatGPT reported 900 million+ weekly active users, 50 million+ subscribers, and 4 million active Codex users; the company closed a record $122 billion funding round at an $852 billion valuation led by Amazon, SoftBank, and Nvidia, filed for an IPO on June 8, 2026, and restructured its Microsoft partnership from exclusive cloud arrangement to a non-exclusive agreement allowing multi-cloud deployment.

What it's good at

Agentic coding (Codex / GPT-5.5)

The Codex agent — powered by GPT-5.5 — executes multi-step software engineering tasks in parallel cloud sandboxes: writing features, fixing bugs, running test suites, and proposing pull requests, with verifiable terminal-log citations. As of April 2026, OpenAI reported 4 million active Codex users.

Integrated chain-of-thought reasoning (GPT-5.5 Thinking)

GPT-5.5's Thinking mode produces an internal reasoning trace before answering; OpenAI reported benchmark scores of 82.7% on Terminal-Bench 2.0 and 51.7% on FrontierMath Tier 1–3. GPT-5 (with thinking) achieves results comparable to standalone o3 while using 50–80% fewer output tokens.

Long-context document processing (1M-token context window)

GPT-5.5 offers a 1 million token context window in the API, enabling full-codebase analysis, long-form legal or financial documents, and multi-document synthesis in a single prompt without chunking.

Native image generation (GPT Image 2 / gpt-image-2)

Launched April 21, 2026, gpt-image-2 achieves approximately 99% character-level text accuracy in Latin, CJK, Hindi, Bengali, and Arabic scripts inside generated images; supports up to 4K resolution, multi-turn context-aware editing, and up to 8 coherent images per prompt with Thinking mode enabled.

Live web search with citations (ChatGPT Search)

ChatGPT Search, launched October 31, 2024, integrates real-time indexed web results with source links directly inside conversations, enabling up-to-date factual queries and positioning ChatGPT as a direct competitor to traditional search engines.

Deep Research (multi-source synthesis)

Deep Research (launched February 2025) autonomously conducts extensive web research and compiles structured multi-source reports in 5–30 minutes, initially powered by o3; the feature is available to Plus and Pro users.

Voice interaction (ChatGPT Voice)

ChatGPT Voice — which replaced Advanced Voice Mode in August 2025 — delivers low-latency natural speech conversation across multiple languages via the mobile and desktop apps; free users can also call ChatGPT by telephone for up to 15 minutes per month, introduced December 2024.

Computer use and GUI automation

GPT-5.5 can autonomously navigate web browsers, operate desktop GUI applications, fill forms, and manipulate spreadsheets, enabling multi-tool agentic workflows without continuous human steering — a capability OpenAI highlighted at GPT-5.5's launch as core to knowledge work.

Backlash & criticism

CEO firing and governance crisis (November 2023)

OpenAI's board abruptly removed Sam Altman on November 17, 2023, citing a failure to be 'consistently candid'; he was reinstated five days later under pressure from Microsoft and employees. Former board member Helen Toner later alleged Altman had withheld information from the board and that two executives reported psychological abuse, raising lasting questions about OpenAI's governance.

Copyright lawsuits (NYT and authors)

The New York Times sued OpenAI and Microsoft in December 2023, alleging unlawful use of millions of copyrighted articles to train ChatGPT; a May 2025 court preservation order required OpenAI to retain all ChatGPT conversation logs affecting 400+ million users worldwide. Numerous authors have filed parallel copyright actions, and the litigation remains ongoing.

Safety researcher exodus (2023–2024)

Roughly half of OpenAI's then-employed AI safety researchers departed between 2023 and 2024, citing concerns about the company's shift toward commercial priorities over safety goals. High-profile departures included Chief Scientist Ilya Sutskever and Alignment team co-lead Jan Leike, who said safety culture had eroded.

GDPR and regulatory bans

Italy banned ChatGPT in March 2023 for collecting personal data without adequate legal basis and failing to provide age verification; service resumed in April 2023 after OpenAI implemented privacy disclosures and opt-out tools. The episode previewed broader EU regulatory scrutiny of ChatGPT's data practices.

Hallucinations and real-world defamation

ChatGPT has generated verifiably false statements about real individuals; in April 2023, an Australian mayor threatened legal action after the chatbot incorrectly stated he had been imprisoned for bribery when he had in fact been a whistleblower. OpenAI has reported reduced but non-zero hallucination rates in successive models.

Child safety wrongful death lawsuit (Raine v. OpenAI)

In August 2025, the parents of a 16-year-old who died by suicide filed a first-of-its-kind wrongful death lawsuit against OpenAI, alleging that months of ChatGPT conversations about mental health and self-harm contributed to the death and that safeguards for minors were inadequate. OpenAI expressed condolences and announced updated crisis-response behaviors and parental controls.

Release timeline

Nov 2022 Jun 2026
  1. Apr 23, 2026
    GPT-5.5 current

    Current flagship; 1M-token context window, strongest agentic coding and scientific reasoning benchmarks to date, matched GPT-5.4 per-token latency; GPT-5.5 Instant became ChatGPT default on May 5, 2026

  2. Mar 1, 2026
    GPT-5.4

    Integrated Codex-trained agentic coding capabilities and native computer-use into the general-purpose model; GPT-5.3 Instant shipped alongside it

  3. Dec 1, 2025
    GPT-5.2

    Improved professional knowledge work, spreadsheet creation, and tool use; shipped Instant, Thinking, and Pro sub-variants

  4. Nov 1, 2025
    GPT-5.1

    First iterative GPT-5 update; introduced Instant and Thinking variants and improved response speed and style

  5. Aug 7, 2025
    GPT-5

    First unified model merging fast GPT-series and o-series reasoning behind a real-time router; made default for all ChatGPT tiers including free users

  6. Dec 1, 2024
    o1 + ChatGPT Pro

    Full o1 reasoning model launched; ChatGPT Pro tier ($200/month) introduced simultaneously alongside the Sora text-to-video model

  7. Sep 12, 2024
    o1-preview

    First public model in OpenAI's reasoning series; introduced the 'thinking before answering' chain-of-thought paradigm

  8. May 13, 2024
    GPT-4o

    Omni model natively processing text, audio, and vision in a single unified model with real-time voice and vision interaction

  9. Nov 6, 2023
    GPT-4 Turbo

    Announced at first OpenAI DevDay; introduced 128K-token context window, JSON mode, and the Assistants API

  10. Mar 14, 2023
    GPT-4

    First multimodal model accepting text and image input; launched simultaneously in ChatGPT and Microsoft Bing with significantly stronger reasoning

  11. Nov 30, 2022
    ChatGPT (GPT-3.5)

    Public research preview on GPT-3.5; reached 100M users in two months — fastest-growing consumer app in internet history at the time