AI models explained for you. Confused by all the buzz around ChatGPT, Microsoft Copilot, Google Gemini, Claude and other AI-branded tools? Who’s building the underlying models, which ones power which services, and how they differ in technology and real-world strengths.
They are all part of the same general AI landscape, but they’re not all the same kind of thing:
- OpenAI and Anthropic are AI labs/companies that build foundation models (GPT, Claude).
- Copilot and Gemini are product brands from Microsoft and Google that package models (often from those labs) into assistants inside apps and services.
Which AI model matters most to you?
Very roughly speaking …
If you live in Microsoft 365 & Windows
Copilot is your most practical layer (because it can see and act on your work), but it’s powered by OpenAI-style models. Knowing how GPT behaves still pays off.
Some people, including us at Office Watch, prefer OpenAI/ChatGPT over Copilot which is more restricted.
If you live in Google Workspace / Android / Search
Gemini will be the most visible; you’ll see it in Gmail, Docs, Sheets, and on your phone.
If you are model-shopping (developer / enterprise)
- OpenAI: strongest general ecosystem and tooling; huge community.
- Anthropic: often chosen where safety/controls and very long-form reasoning are priorities.
- Google Gemini: attractive if you’re already on Google Cloud or want tight integration with search/video/data.
OpenAI (ChatGPT / GPT models)
OpenAI is an AI research and product company. Its public face is ChatGPT, powered by its GPT model family (currently GPT-4.1, GPT-5, GPT-5.1, etc.).
General-purpose, top-tier reasoning and multimodal capabilities, plus a big ecosystem (custom GPTs, plugins, etc.). It’s both a consumer product and the “engine room” for lots of other services.
Where you see it
- ChatGPT website and apps.
- APIs used by developers.
- Under the hood in many third-party tools (and in parts of Microsoft Copilot).
Tech & strengths
- GPT-5 / GPT-5.1 are marketed as OpenAI’s “smartest, fastest, most useful” models with built-in “thinking” for more complex reasoning.
- GPT-4.1 and later models offer 1M-token context windows, strong coding and reasoning, and tool-calling.
- ChatGPT has ~800M weekly active users as of late 2025, making it by far the best-known general AI assistant.
Microsoft Copilot
Copilot is Microsoft’s brand for its AI assistants across Windows, Microsoft 365, GitHub, Edge, Bing, and “Copilot+ PCs.” It’s not an AI model itself; it’s a product layer that sits on top of models (primarily OpenAI GPT-4/5 plus some Microsoft models).
“Your work Copilot.” Its main differentiator isn’t the raw model (which is often OpenAI under the hood), but the tight integration with Microsoft 365 and Windows and the ability to act on your files and enterprise data.
Where you see it
- Microsoft 365 Copilot inside Word, Excel, PowerPoint, Outlook, Teams, etc.
- Windows Copilot / Copilot+ PCs – assistants integrated into the OS, with new NPU-heavy hardware.
- GitHub Copilot for coding.
- Copilot web/mobile apps and Bing.
Tech & strengths
- Uses OpenAI models (GPT-4, GPT-4.1, GPT-5 family) plus Microsoft’s own components.
- Microsoft appears to limit OpenAI responses, seemingly to avoid controversy.
- Deeply integrated with Microsoft Graph and organization’s data – documents, emails, calendar, Teams chats – with permissions respected.
- Newer features like Agent Mode (Word/Excel) and Copilot Actions let it do things for you (organize files, process documents) in an isolated “agent workspace” on your PC.
Google Gemini
Gemini is Google’s family of large language models and the consumer assistant brand. It replaced Bard and now powers the Gemini chat app, Google Search’s AI answers, Workspace features, Android, and various Google Cloud services.
Google’s answer to ChatGPT + Copilot: a general assistant but highly integrated into Google Search and Workspace, plus broad deployment across Android and Google Cloud.
Where you see it
- Gemini app / web (chat assistant).
- Google Search’s AI overviews.
- Gmail, Docs, Sheets, Slides (Workspace).
- Android (Gemini on-device), Chrome, and via Vertex AI on Google Cloud.
Tech & strengths
- A family of models: previously Gemini 1.x and 2.5 (Pro, Flash, Flash-Lite), and now Gemini 3.0 Pro / Deep Think as Google’s top-end reasoning models.
- Strong focus on multimodality (text, images, video, audio in one context) and very long context windows (up to 1M tokens in some variants).
- Tight coupling with Google’s ecosystem: Search, Maps, YouTube, Gmail, Docs, etc.
Anthropic (Claude)
Anthropic is an AI company founded by ex-OpenAI researchers. Its flagship product is Claude, a family of models (Claude 3, 3.5, 4.x series like Claude Opus 4.1).
A general-purpose assistant like ChatGPT/Gemini, but marketed heavily on safety, controllability, and enterprise-grade reliability, with strong traction via cloud platforms and SaaS integrations.
Where you see it
- Claude.ai web app and mobile.
- APIs via Anthropic, and integrations via Amazon Bedrock, Google Cloud Vertex AI, and other platforms.
- It’s also available as an option for some Microsoft AI services, for example in Excel.
Tech & strengths
- Claude 3.5 Sonnet set a new bar for intelligence/cost in mid-2024, outperforming previous Claude 3 Opus in many benchmarks while being faster and cheaper.
- Anthropic is strongly associated with “Constitutional AI” – training methods that align the model to a written set of principles and use AI feedback to improve harmlessness and steerability.
- Known for long context, good writing/analysis, and conservative/safety-forward behavior.
How the AI models relate
A few important relationships:
Gemini vs OpenAI
OpenAI+Microsoft is one axis, Google+Gemini is the other: same general idea (assistant everywhere, in apps, search, cloud).
Copilot often runs on OpenAI models
Microsoft Copilot (especially Microsoft 365 Copilot and Windows Copilot) mostly uses GPT-4/5 under the hood, combined with Microsoft’s own orchestration and data-access layers.
So “using Copilot” often means you’re indirectly using OpenAI tech, but with Microsoft handling the integration, security and licensing.
Anthropic/Claude is often the “third option” for enterprises.
Many businesses use Claude via Amazon Bedrock or Vertex AI alongside GPT and Gemini, especially where safety and interpretability are big selling points.
Everyone is adjusting limits & pricing as usage explodes
OpenAI and Google have both trimmed free usage of heavy models (e.g., Sora 2 video, Gemini Pro, “Nano Banana Pro” image gen) due to GPU costs, nudging users to paid tiers.
Meanwhile, Microsoft is pushing Copilot hard, perhaps too hard. Elements of Copilot are available to many Microsoft 365 subscribers.
Discover the COPILOT Function in Excel
How many Copilot AI Credits do I have?
Microsoft 365 Premium: Copilot AI, new Features & Pricing Explained
