ChatGPTGeminiClaudeGrok40,000+users on the Chrome Web Store4.7/ 5

ChatGPT Glossary: 56 AI Terms, Defined Plainly

Short definitions of the words that come up around ChatGPT and the other assistants: models and their retirement dates, tokens and context windows, prompts, RLHF, hallucination, and the organizing features that ChatGPT lacks natively. Each entry ends with the practical detail, and each has its own link.

Add to Chrome, freeStart at A

Glossary free forever | Context gauge and full-text search free inside ChatGPT, Gemini, Claude and Grok | Premium $9.99/mo or $99 once

Jump to a letter

All 56 terms

A to V

The terms

Alphabetical. Terms that describe an AI Toolbox feature link to the page that documents it.

A

2 terms
Agent / Agentic AI
An AI system that can take autonomous actions, make decisions, and execute multi-step tasks without human intervention. OpenAI's Agents SDK and ChatGPT's agent capabilities enable automated workflows.
API (Application Programming Interface)
A programming interface that lets developers access ChatGPT programmatically. OpenAI's API supports GPT-6 Astra, the GPT-5.6 family (Sol, Terra, Luna), GPT-5.5, GPT-5.4, the GPT Image 2 image model, and other current models with usage-based pricing.

B

3 terms
Bulk Archive
Archiving multiple ChatGPT conversations simultaneously. ChatGPT archives one conversation at a time; AI Toolbox adds multi-select archive in Manage Chats, 5 chats per batch on the free plan and unlimited on Premium.
Bulk Delete
Deleting multiple ChatGPT conversations at once. ChatGPT's native interface requires deleting one chat at a time. AI Toolbox adds checkbox selection and bulk delete in Manage Chats, 5 chats per batch on the free plan and unlimited on Premium.
Bulk Export
Exporting multiple ChatGPT conversations at once as TXT, Markdown, JSON or PDF. Single-chat TXT export is free in AI Toolbox; bulk export and the other formats are Premium. Used for backups, documentation and knowledge management.

C

10 terms
ChatGPT
OpenAI's conversational AI chatbot, launched November 2022. The most widely used AI assistant with hundreds of millions of users, available as a web app at chatgpt.com and mobile apps for iOS and Android.
ChatGPT Free
ChatGPT's free tier providing access to a default everyday model with limited message quotas (approximately 10 messages per 5 hours for advanced models). Includes basic features like chat, file uploads, and web search.
ChatGPT Plus
A $20/month ChatGPT subscription with GPT-5.6 Sol and its Medium and High thinking levels in chat, GPT-6 Astra in ChatGPT Work and Codex, a 54K-token context window, Projects, deep research, GPT Image 2 image generation and voice. OpenAI publishes its usage limits as multiples of the Free allowance rather than as fixed message counts.
ChatGPT Pro
ChatGPT's top individual plan at $100/month for 5x the usage of Plus or $200/month for 20x. It adds Pro reasoning powered by GPT-6 Astra, the highest reasoning effort, maximum Codex and deep research allowances, and a 128K (instant) to 400K (reasoning) context window. Designed for researchers and power users who need maximum compute.
ChatGPT Team
A $25/user/month ChatGPT plan for workplace collaboration with admin controls, shared workspace, SAML SSO, and higher usage limits than Plus. Minimum 2 seats.
ChatGPT Toolbox
The former name of AI Toolbox. One install with four togglable modules, for ChatGPT, Gemini, Claude and Grok, that adds folders, full-text search, bulk export, a prompt library and prompt chains inside each. Runs on Chrome, Edge, Brave, Opera, Arc and other Chromium browsers, with a free plan.
Chrome Extension
Software that adds functionality to the Google Chrome browser. ChatGPT extensions such as AI Toolbox, AIPRM and Superpower ChatGPT add features to ChatGPT's interface. AI Toolbox runs on every Chromium browser (Chrome, Edge, Brave, Opera, Arc); Firefox is not supported.
Context Window
The maximum amount of text (measured in tokens) a model can process in a single conversation. On the API, GPT-6 Astra, GPT-5.6, GPT-5.5 and GPT-5.4 support a 1.05 million token context window (about 750,000 words); inside ChatGPT the window follows the plan, 27K on Free, 54K on Go and Plus, and 128K to 400K on Pro. Once a conversation exceeds the window, the earliest turns drop out of what the model can see; the AI Toolbox Context Meter shows how much of the window is used.
Custom GPT
User-created specialized versions of ChatGPT with custom instructions, knowledge files, and optional API actions, built in the GPT Builder. Personal plans (Free, Go, Plus, Pro) can no longer create new GPTs; Business, Enterprise, and Edu workspaces still can where settings allow. OpenAI plans to retire custom GPTs on December 11, 2026 and move them to plugins.
Custom Instructions
Persistent preferences saved in ChatGPT settings that shape how the AI responds to all your conversations. Includes 'What would you like ChatGPT to know about you?' and 'How would you like ChatGPT to respond?'

D

2 terms
DALL-E (deprecated)
OpenAI's earlier image generation model family (DALL-E 2, DALL-E 3) that was integrated into ChatGPT. Now retired. ChatGPT image generation is currently powered by GPT Image 2, with earlier image models still surfaced in the AI Toolbox Media Gallery for historical content.
Deep Research
A ChatGPT feature that performs multi-step web research, synthesizes information from multiple sources, and generates comprehensive reports with citations. Available to Plus and Pro users.

F

2 terms
Fine-tuning
Training an AI model on specific data to improve performance for a particular domain or task. OpenAI offers fine-tuning for select current models through their API.
Folders
An organizational system for grouping ChatGPT conversations by topic, project, or category. ChatGPT has Projects, a single level with no nesting. AI Toolbox adds a My Folders tree to the sidebar, filed by search or Move to Folder: 2 root folders with unlimited subfolders free, unlimited on Premium.

G

10 terms
GPT (Generative Pre-trained Transformer)
The neural network architecture behind ChatGPT, developed by OpenAI. GPT models are pre-trained on large text datasets and fine-tuned with RLHF to follow instructions and have conversations.
GPT-4 (retired)
OpenAI's fourth-generation large language model, released March 2023. Retired by OpenAI in 2026 and superseded by the GPT-5.x family. Current users should use GPT-6 Astra, GPT-5.6 Sol, or GPT-5.5 instead.
GPT-4o (retired)
An optimized version of GPT-4 ('o' for 'omni') released May 2024. Retired by OpenAI on February 13, 2026. Current users should use GPT-6 Astra, GPT-5.6 Sol, or GPT-5.5 instead.
GPT-4.1 (superseded)
An earlier OpenAI ChatGPT model. Superseded by the GPT-5.x and GPT-5.6 families and no longer part of OpenAI's frontier lineup as of the GPT-5.6 launch.
GPT-5 (superseded)
An OpenAI ChatGPT model in the original GPT-5 family. Superseded by later generations (GPT-5.4, GPT-5.5, the GPT-5.6 family and GPT-6 Astra); OpenAI's model page now recommends GPT-6 Astra instead.
GPT-5.4
An OpenAI ChatGPT model released March 2026, available with Pro, mini, and nano variants. Sits below GPT-5.5, the GPT-5.6 family and GPT-6 Astra in capability.
GPT-5.5
A high-tier OpenAI ChatGPT model available with a Pro variant for maximum compute. Sits just below the GPT-5.6 family (Sol, Terra, Luna), which in turn sits below GPT-6 Astra.
GPT-5.6 (Sol, Terra, Luna)
OpenAI's frontier model family launched July 2026, one step below GPT-6 Astra since September 4, 2026. GPT-5.6 Sol is the family's flagship for complex professional work, GPT-5.6 Terra balances intelligence and cost, and GPT-5.6 Luna targets cost-sensitive, high-volume workloads. All three share a ~1.05M-token context window and add a new max reasoning effort plus an ultra subagent mode.
GPT-6 Astra
OpenAI's flagship model, announced September 3, 2026 and rolling out from September 4 to ChatGPT Plus, Pro, Business and Enterprise, and to the API. It carries a 1.05 million token API context window, 128K max output, an April 30, 2026 knowledge cutoff, and API pricing of $10 input and $50 output per million tokens. Free and Go accounts stay on GPT-5.6 Luna.
GPT Image 2
OpenAI's current image generation model live on ChatGPT in 2026. Replaced earlier image models (GPT Image 1.5, GPT Image 1, DALL-E 3, DALL-E 2). The AI Toolbox Media Gallery organizes images from GPT Image 2 plus earlier image models.

H

1 term
Hallucination
When an AI model generates plausible-sounding but factually incorrect information. ChatGPT can hallucinate citations, statistics, dates, and events. Always verify critical facts from AI responses.

J

1 term
JSON Export
Exporting ChatGPT conversations in structured JSON format, preserving message metadata, timestamps, and conversation structure. An AI Toolbox feature useful for developers and data analysis.

L

1 term
Large Language Model (LLM)
An AI model trained on vast amounts of text data to understand and generate human language. Examples include GPT-6 Astra (OpenAI), Claude (Anthropic), Gemini (Google), and Llama (Meta).

M

4 terms
Media Gallery
An AI Toolbox feature that lets you browse, download, and manage all AI-generated images from ChatGPT (GPT Image 2 live on ChatGPT, plus earlier image models). View prompts, Gen IDs, Seed IDs, and bulk download as ZIP.
Memory
ChatGPT's ability to remember information across separate conversations. When enabled, ChatGPT stores facts you share (preferences, work context, personal details) and applies them automatically.
Model
The underlying AI system that processes your prompts. ChatGPT offers multiple current models: GPT-6 Astra, GPT-5.6 Sol, GPT-5.6 Terra, GPT-5.6 Luna, GPT-5.5, GPT-5.5 Pro, GPT-5.4, GPT-5.4 Pro, GPT-5.4 mini, GPT-5.4 nano.
Multimodal
An AI model capable of processing multiple input types (text, images, audio and video) in a single interaction. The current GPT-5.x family includes multimodal models.

O

4 terms
o1 (retired)
OpenAI's first reasoning-focused model, released September 2024. Retired by OpenAI and superseded by the GPT-5.x family. Current reasoning workloads should use GPT-6 Astra, GPT-5.6 Sol, or GPT-5.5 Pro.
o3 (retired)
OpenAI's reasoning model released April 2025. Retired by OpenAI on February 13, 2026. Current reasoning workloads should use GPT-6 Astra, GPT-5.6 Sol, or GPT-5.5 Pro.
o4-mini (retired)
A fast, efficient reasoning model released April 2025. Retired by OpenAI on February 13, 2026. Current users should use GPT-5.4 mini or GPT-5.6 Luna for similar speed/cost balance.
OpenAI
The artificial intelligence company that created ChatGPT, the GPT model family, GPT Image 2, Whisper, and Sora. Founded in 2015 by Sam Altman, Greg Brockman, and others. Headquartered in San Francisco.

P

5 terms
Pin Conversations
Keeping important ChatGPT chats fixed at the top of your sidebar for instant access. An AI Toolbox feature, since ChatGPT's native interface does not support pinning conversations: 2 pinned chats free, unlimited on Premium.
Prompt
The text input you give to ChatGPT. A prompt can be a question, instruction, creative brief, or any text that tells the AI what to do. Better prompts produce better responses.
Prompt Chaining
Running multiple prompts sequentially where the output of one feeds into the next, with dynamic variables. An AI Toolbox Premium feature, triggered by typing two dots in the compose box, for automating multi-step AI workflows.
Prompt Engineering
The practice of crafting effective prompts to get better, more accurate, and more useful responses from AI models. Techniques include role-setting, few-shot examples, chain-of-thought, and structured formatting.
Prompt Library
A collection of saved, reusable prompts. AI Toolbox includes a built-in Prompt Library where you save prompts with dynamic {{fields}} and insert them by typing two slashes in the compose box: 2 saved prompts and 5 catalog prompts free, unlimited on Premium.

R

4 terms
RAG (Retrieval-Augmented Generation)
A technique that combines AI text generation with real-time data retrieval from external sources. Used by ChatGPT's web browsing feature and Deep Research to provide up-to-date, cited answers.
Rate Limit
The maximum number of messages or API calls allowed per time period. ChatGPT Free allows ~10 messages per 5 hours for advanced models. Plus allows ~160 messages per 3 hours on current GPT-5.x models.
RLHF (Reinforcement Learning from Human Feedback)
A training technique where human reviewers rank AI outputs and the model learns to produce higher-ranked responses. RLHF is the primary method OpenAI uses to align ChatGPT with human preferences.
RTL (Right-to-Left)
Text direction used by Arabic, Hebrew, and other scripts. AI Toolbox provides full RTL layout for Arabic and Hebrew across a 10+ language UI with automatic text direction detection.

S

3 terms
Sora
OpenAI's video generation model that creates realistic video from text descriptions. Integrated into ChatGPT for Plus and Pro subscribers.
Subfolders
Nested folders within folders for deeper organizational hierarchy. An AI Toolbox feature with unlimited nesting on every plan, so conversations can be organized by project, topic and subtopic.
System Prompt
Hidden instructions that define ChatGPT's behavior, personality, and constraints in a conversation. Custom GPTs use system prompts to create specialized assistants. Also called 'system message'.

T

3 terms
Temperature
A parameter (0-2) controlling the randomness of AI responses. Temperature 0 gives the most deterministic output. Temperature 1+ increases creativity and variation. Default is typically 0.7-1.0.
Token
The basic unit of text processed by AI models. One token is roughly 4 English characters or 0.75 words. Pricing varies by model, see openai.com/pricing for current rates.
TXT Export
Exporting ChatGPT conversations as plain text files for easy reading, sharing, and printing. Single-chat TXT export is free in AI Toolbox; bulk TXT export is Premium.

V

1 term
Voice Mode
ChatGPT's spoken conversation feature enabling real-time audio interaction. Advanced Voice Mode (available to Plus/Pro) offers natural, expressive speech with multiple voice options. AI Toolbox can download ChatGPT voice replies as MP3 files.

Using the glossary

Find, read, follow, share

Free, no account, and every definition is a plain-language paragraph you can link to.

  • Jump by letter or search

    The letter row scrolls to a group; the search box matches term names and definitions, so "token" or "RLHF" lands on the entry without scrolling.

  • Read the definition, then the caveat

    Most entries end with the practical detail: which plan a feature is on, when a model was retired, what to check before relying on it.

  • Follow the link where there is one

    Terms that describe an AI Toolbox feature link to the page that documents it, with captures and the free versus Premium split.

  • Send a term to someone

    Every term has its own anchor. Copy the link from the entry and it opens the page scrolled to that definition.

In the chat itself

Two of these terms, as things you can see

This page defines the words. AI Toolbox puts the numbers behind them inside ChatGPT, Gemini, Claude and Grok.

Context window

See the number, not just the definition

A context window is a figure on a spec sheet until a long thread starts forgetting its start. The AI Toolbox Context Meter reads the running token count inside ChatGPT, Gemini, Claude and Grok, warns when the window is nearly full, and can hand the thread to a fresh chat with a summary. The gauge is free on all four.

Plan: Free: the gauge on all four platforms. Premium: cross-platform handoff.

Context Meter panel reading 479k of 500k tokens with a context window is nearly full warning, about 9 messages left, and a Continue in a fresh chat handoff

Search

Find the chat by what it said

ChatGPT's own search matches titles. AI Toolbox full-text search matches every message, filters by role, date or exact phrase, and opens the conversation at the matching turn, which is how you find the thread where a term was first explained to you.

Plan: Free: 5 results per query. Premium: unlimited, with advanced filters.

Search History panel showing 'launch' matched in 11 messages across 3 conversations, with per-message highlights and bulk actions

Pricing

The glossary is free. So is a lot of the extension.

The context gauge, 2 root folders with unlimited subfolders, 2 pins and 5 search results per query are on the free plan. Premium removes the caps and adds bulk export. Not an AI subscription: you keep using your own accounts.

Free

$0forever

  • Context gauge on all four platforms
  • 2 root folders, unlimited subfolders
  • 5 search results per query
  • 2 saved prompts inserted with two slashes
Add to Chrome

Premium

$9.99/month

  • Unlimited folders, pins and search
  • Bulk archive, delete and export
  • Unlimited prompts and prompt chains
  • Cross-platform context handoff
Get Premium

LifetimeMost popular

$99one-time

  • Everything in Premium
  • Pay once, use forever
  • All future updates
  • Top priority support

Pays back in under 10 months of Premium.

Get Lifetime Access

All Access LifetimeBest value

$199one-time

  • ChatGPT, Gemini, Claude and Grok included
  • Every future AI platform we add
  • All Premium features unlocked
  • Save $197 vs single Lifetimes
Get All Access

Teams$15/seat/month

All Premium features across the four modules · Admin dashboard · Team analytics · $12/seat/mo billed annually · 14-day free trial

See Teams
  • GDPR compliant
  • Secure checkout by Polar
  • 14-day money-back on a first purchase
  • Chats stay on the AI platform you use

Refunds follow the refund policy; renewals are not refundable. Premium unlocks for the email you enter at checkout, so use the same address you sign in to ChatGPT with.

FAQ

The questions behind the terms

What is ChatGPT?

ChatGPT is OpenAI's conversational AI assistant, launched in November 2022. It runs on the current GPT model lineup (GPT-6 Astra, GPT-5.6 Sol, GPT-5.6 Terra, GPT-5.6 Luna, GPT-5.5, GPT-5.5 Pro, GPT-5.4, GPT-5.4 Pro, GPT-5.4 mini, GPT-5.4 nano) and is available at chatgpt.com and in mobile apps, with a free tier and paid Plus, Pro, Team and Enterprise plans.

What is a token in ChatGPT?

A token is the unit of text a model reads and writes: roughly four English characters, or about three quarters of a word. Context windows, rate limits and API prices are all counted in tokens. The free token counter on this site estimates the count for any pasted text.

What is the difference between a context window and memory?

The context window is what the model can see in the current conversation, measured in tokens; when it fills, the earliest turns drop out. Memory is a separate ChatGPT feature that stores facts about you across conversations and injects them into new ones. A large window does not give the model memory, and memory does not enlarge the window.

What ChatGPT models are available in 2026?

GPT-6 Astra, GPT-5.6 Sol, GPT-5.6 Terra, GPT-5.6 Luna, GPT-5.5, GPT-5.5 Pro, GPT-5.4, GPT-5.4 Pro, GPT-5.4 mini, GPT-5.4 nano. GPT-4o, o3, o4-mini and earlier models were retired on February 13, 2026 and replaced by the GPT-5.x family; GPT-4.1 and the original GPT-5 remain on the API but are no longer frontier models.

What is prompt engineering?

The practice of writing prompts that get better answers: naming a role, stating the task and the output format, adding constraints, giving examples, and asking for reasoning before the answer. The prompt formatter on this site wraps any prompt in those structures.

What is RLHF?

Reinforcement Learning from Human Feedback: human reviewers rank model outputs and the model is trained to prefer the higher-ranked ones. It is the main technique OpenAI uses to make ChatGPT follow instructions and decline harmful requests, on top of the base pre-training.

What is a hallucination, and how do I catch one?

A confident answer that is wrong: an invented citation, a date, a statistic, a function that does not exist. Ask the model for its sources and check them, prefer answers grounded in web search or your own documents, and treat numbers and names as unverified until you have confirmed them.

What is AI Toolbox?

A browser extension for Chromium browsers, formerly ChatGPT Toolbox, with modules for ChatGPT, Gemini, Claude and Grok in one install. It adds folders, full-text search, bulk archive, delete and export, a prompt library with chains, pins, a context gauge and a media gallery. Used by 40,000+ people, rated 4.7/5 on the Chrome Web Store. The free plan covers the basics; Premium is $9.99/month or $99 once per module.

Keep readingChatGPT models guideToken counterAI model comparisonPrompt formatterAI Toolbox ChatGPT featuresAll extension features

40,000+ users · 4.7/5 on the Chrome Web Store

Put the vocabulary to work

Install AI Toolbox for folders, full-text search, a context gauge and a prompt library inside ChatGPT, Gemini, Claude and Grok.

Add to Chrome, freeSee all plans