Claude vs ChatGPT vs Gemini: the three-way comparison
The honest answer to "which one" is "it depends on the message": Claude for careful long-form reasoning and the best memory controls, ChatGPT for the most polished all-round product, Gemini for the million-token window and the bundle. That is precisely the problem. Each app locks its excellent models behind its own walls, so the people who care most end up paying for two or three and ferrying context between them by hand. We build NodePad, and the fourth column is our answer: Claude, GPT, and Gemini in one workspace, picked per message, each seeing exactly the context you assembled. If you only ever need one vendor, buy that vendor. If you find yourself asking "what would the other model say", you are the person this page's fourth column exists for.
| dimension | Claude | ChatGPT | Gemini | NodePad |
|---|---|---|---|---|
| Price of entry | Free; Pro $20/mo ($17 annual); Max $100 to $200/mo. | Free; Go $8/mo; Plus $20/mo; Pro $200/mo. | Free; AI Plus $4.99/mo; AI Pro ~$20/mo; AI Ultra $100 to $200/mo. | Free on community models; Go $5, Pro $20, Max $100. Every paid plan reaches every model below. |
| Models | Haiku, Sonnet 5, Opus 4.8, Fable 5. Anthropic only. | GPT-5.5 free; GPT-5.6 Sol, Terra, Luna paid. OpenAI only. | Fast, Thinking, Pro labels over the Gemini 3.x family. Google only. | Claude, GPT, and Gemini side by side; the model is chosen per message. |
| Context window | 200K tokens on all consumer plans; long chats are auto-summarized. | In-app windows are unpublished and far below the API's (roughly 32K on Plus). | 32K free, 128K Plus, 1M on Pro and Ultra; the biggest paid window here. | The model's full window, spent on the context you assembled instead of a resent scrollback. |
| Who assembles the context | Memory synthesis plus retrieval decide what enters; you can edit the summary, not the per-message context. | Hidden system prompt plus background-rewritten memory. | Personalization from chat history and, opted in, your Google apps. | You. The path behind your message plus what you referenced or merged. Nothing hidden, nothing injected. |
| Branching | Edit and regenerate create hidden versions behind per-message arrows; no view of the tree. | "Branch in new chat" copies context to a separate flat chat. | None; strictly linear. | Fork any point onto a visible parallel path on the canvas; the original continues untouched. |
| Working across chats | Projects share knowledge and instructions; the chats still cannot see each other. | Projects group chats and files; same isolation. | No working folders shipped as of mid-2026. | Multiple chats on one canvas; merge pulls several into scope for one answer. |
| Memory | The best controls anywhere; view, edit, pause, reset, per-project isolation, on every plan. | The deepest recall anywhere, and the least inspectable. | A personalization toggle plus deletable activity. | None by design; continuity is the canvas you can see. |
| Beyond chat | Artifacts, Claude Code bundled at $20, voice, MCP ecosystem. | Best apps, full-duplex voice, image generation, agents, custom GPTs. | Workspace integration, Gemini Live free, image and video generation, the storage bundle. | Chat on a canvas is the whole product; plugins add tools. These three win this row in three different ways. |
| Usage limits | A conversation budget with no published numbers; chat and Claude Code share one pool. | Dynamic and unpublished. | Published as multipliers only, never absolute numbers. | A visible meter; add usage or wait for the reset. No hidden throttles. |
No credit card to start. Every paid plan reaches every model.