Compare
Six AI tools, side by side.
What changes between ChatGPT, Claude, Copilot, Gemini, Perplexity, and NotebookLM, written for someone who just wants to know which one to open Monday morning. Tiers, message limits, default modes, what each one is genuinely best at, and where it tends to fall down.
| Attribute | ChatGPT chat.openai.com | Claude claude.ai | Copilot copilot.microsoft.com | Gemini gemini.google.com | NotebookLM notebooklm.google.com | Perplexity perplexity.ai |
|---|---|---|---|---|---|---|
| Vendor | OpenAI · San Francisco AI company. Makers of GPT. | Anthropic · San Francisco AI company. Makers of Claude. | Microsoft · The company behind Windows and Office. | Google · The company behind Search, Gmail, and Android. | Google · The company behind Search, Gmail, and Android. | Perplexity · San Francisco AI company. Search-first chat. |
| Tiers | Free · Plus · Pro · Business · Enterprise | Free · Pro · Max · Team · Enterprise | Free · Pro · M365 Copilot (paid add-on) | Free · AI Pro · AI Ultra · Workspace add-on | Free · Plus | Free · Pro · Enterprise Pro |
| We tested on | Plus (paid personal) | claude.ai default | M365 Copilot (paid personal subscription) | Gemini 3.5 Flash and Flash Extended | Free tier | Auto (default routing) |
| Default model | GPT-5.5 Instant | Claude Sonnet 4.6 | GPT family (Microsoft-tuned) | Gemini 3.5 Flash (Free) · Gemini 3 Pro (Pro) | Gemini family (Google-tuned for sources) | Auto / Auto-Max (model auto-selected) |
| Smart mode for hard work | GPT-5.5 Thinking | Opus 4 (paid) | Think Deeper | Gemini 3 Pro (paid) | Audio Overview, Study Guide, Mind Map (built-in) | Reasoning models (paid) |
| Message length cap | No published character cap; long pastes accepted on paid tiers. | Long pastes accepted. Soft caps on the free tier. | Free: 10,240 characters per message (hard cap). Paid: much higher. | Long pastes accepted on all tiers. | Generous chat length. Limits sit on file uploads, not message size. | Long pastes accepted. |
| Context window | ~32k tokens (Free) · ~128k+ tokens (Plus and Pro) | ~200k tokens on all tiers | Not publicly disclosed; effective limit governed by the message cap. | ~1M tokens on paid tiers (largest in this group) | Holds entire notebook of uploaded files (~50 sources on Free; ~300 on Plus). | Varies by routed model; long enough for full transcripts. |
| Daily limits | Free: tight daily limits. Plus: generous. Pro: highest. | Free: usable for short bursts. Pro: generous all day. | Free: limited daily turns. Pro: relaxed. M365 Copilot: included in subscription. | Free: limited Pro access per day. AI Pro: generous. | Free: limited daily Audio Overviews. Plus: higher caps. | Free: limited Pro searches per day. Pro: 300+ Pro searches per day. |
| Files, web, integrations | Files: yes. Web search: yes. Image input: yes. | Files: yes. Web search: yes. Image input: yes. Code in Artifacts. | Files: yes (paid). Web search: yes. Lives inside Word, Excel, Outlook, Teams. | Files: yes. Web search: native. Direct hooks into Gmail, Docs, Drive, Meet. | Files: yes (PDF, Docs, Slides, YouTube, web pages, audio). Web search: no. | Files: yes. Web search: native (every answer is a search). |
| Best at | Drafting, brainstorming, structured outputs, coding help. | Long-document reading, careful writing, executive summaries. | Working directly inside Office documents and Teams meetings. | Reading Google Meet transcripts, working across Gmail and Docs. | Answering only from your uploaded documents. Refuses to invent answers. | Answers that need real, citable sources. Live research. |
| Watch out for | Default Instant mode sometimes overshoots word caps. Ask it to count. | No native image generation. Pair with another tool for visuals. | The free message cap blocks real meeting transcripts. Step up to a paid tier. | Flash sometimes misses items buried in long transcripts. Step up for high-stakes work. | Cannot answer about anything outside the notebook. By design. | Source-first behavior means short questions can return long, link-heavy answers. |
Notes for the curious
- Tokens vs. characters. A token is about 4 characters or three-quarters of an English word. A 200k-token context window holds roughly 150,000 words at once. See the Glossary for the long version.
- Why the modes differ. Every product offers fast modes and slow modes because the underlying models trade speed for depth. Pick fast for everyday work, slow for the answers that need to be right. See the Glossary for each mode written out.
- Free Copilot has a hard 10,240-character cap per message. A 60-minute meeting transcript will not fit. For our Issue 01 status-report test, we used M365 Copilot on a paid personal subscription.
- Limits change. Tiers, prices, model names, message caps, daily limits, and context windows all shift without warning. A vendor can rename a tier overnight, swap which model is the default, raise or lower the message cap, or quietly retire a mode. We re-test before every Monday issue and update this page. Anything you see here was true the week we published. Last verified June 2026.