ChatGPT vs Claude.
OpenAI's assistant against Anthropic's: the current models, what each can read and make, what the plans and the API cost, and which to pick for which job.
ChatGPT
ClaudeFeature by feature.
Models, as of November 2026
| Feature | ChatGPT | Claude |
|---|---|---|
| Top model | GPT-6 Astra | Claude Fable 5.2 |
| On the free plan | GPT-6 Luna | Claude Sonnet and Haiku |
| Paid plans add | GPT-6 Sol, GPT-6.1 Sol and GPT-6 Astra | Claude Opus 5.5 and Fable 5.2 |
Reading & Making
| Feature | ChatGPT | Claude |
|---|---|---|
| Reads (API) | Both Text, images and PDFs | |
| File uploads in the app | Both Yes | |
| Makes images | Yes, in ChatGPT and the API | No; diagrams and charts in code |
| Voice | Voice chats, and voice with video | Voice mode, in beta |
| Web search | Both Built in | |
| Deep research | Limited on Free, more on paid plans | On paid plans |
Work & Tools
| Feature | ChatGPT | Claude |
|---|---|---|
| Coding agent | Codex | Claude Code, on paid plans |
| Working in a browser | A built-in browser | Claude in Chrome, on paid plans |
| Shared assistants | Custom GPTs, Plus and up | Projects and Skills |
| Office apps | Excel, Word, PowerPoint and Google Sheets | Microsoft 365, on paid plans |
| Memory of past chats | Limited on Free, more on paid plans | On every plan |
API, per million tokens
| Feature | ChatGPT | Claude |
|---|---|---|
| Context window | 1.05 million tokens | 1 million tokens |
| Longest reply | Both 128,000 tokens | |
| Top model | $10 in, $50 out (GPT-6 Astra) | $8 in, $40 out (Fable 5.2) |
| Mid model | $2 in, $10 out (GPT-6.1 Sol) | $2 in, $10 out (Sonnet 5.5) |
| Smallest model | $0.10 in, $0.50 out (GPT-6 Luna) | $0.10 in, $0.50 out (Haiku 5.5, prompts to 100,000 tokens) |
| Long prompts | Twice the input price above 272,000 tokens | Standard price to 1 million tokens, Haiku aside |
| Batch requests | Both Half price | |
Privacy
| Feature | ChatGPT | Claude |
|---|---|---|
| Consumer chats used for training | Yes, unless you opt out | Your choice, in privacy settings |
| Business plans | Both No training by default | |
| API | Both Not used for training by default | |
Plans & pricing.
| Plan | ChatGPT | Claude |
|---|---|---|
| Free | $0 | $0 |
| Entry plan | Go: $6/month, may include ads | None between Free and Pro |
| Standard plan | Plus: $20/month | Pro: $20/month, or $17/month billed yearly |
| Heavy use | Pro: $100, $200 or $500/month | Max: $100/month; Ultra: $200/month |
| Team seat | $20/seat billed yearly, $25 monthly | $20/seat billed yearly, $25 monthly |
| Premium team seat | $100/seat billed yearly, $125 monthly | $100/seat billed yearly, $125 monthly |
| Enterprise | Custom pricing | $20/seat/month billed yearly plus API usage, 20 seats minimum |
US dollars as of November 2026, from chatgpt.com/pricing, claude.com/pricing and each company's API pricing page. Prices can differ by country.
Our conclusion.
Pick ChatGPT when you want images made in the same app, the cheapest paid plan (Go, $6 a month) or custom GPTs your team can share. Pick Claude for long prompts through the API, which Anthropic bills at its standard rate up to 1 million tokens, or when you want to choose whether your chats train its models.
For writing, research and code, neither company's published results settle which is better for your work, so the clearer differences are the tools. ChatGPT makes images and Claude does not; ChatGPT has a $6 plan (which may include ads) and Claude has none between Free and Pro; ChatGPT's consumer plans train on your chats unless you opt out, while Claude asks you to choose.
For a team the price is the same: $20 a seat a month billed yearly on ChatGPT Business and Claude Team, and neither trains on business data by default. Through the API, OpenAI's top model costs $10 in and $50 out per million tokens, Anthropic's $8 and $40, and their mid and small models cost the same as each other.
Models change often; we review this page every quarter.
Where we come in.
- AI ChatbotsTrained on your data, integrated with your systems, and designed to escalate when they should. Not frustrate your customers.See the page
- AI AgentsAutonomous agents that handle research, outreach, scheduling, and analysis. Working alongside your team.See the page
- AI Business ToolsCustom dashboards, content generators, and decision support systems powered by AI. Designed for how your team actually works.See the page
- AIWe build chatbots, automation, agents, and internal tools around how your business actually works, eliminating repetitive tasks and freeing your team to focus on what matters.See the page
Other comparisons.
ChatGPT vs Gemini
Compare ChatGPT and Gemini on models, tools, plans, API prices and privacy.
See ChatGPT vs GeminiClaude vs Gemini
Compare Claude and Gemini on models, tools, plans, API prices and privacy.
See Claude vs Gemini
GeminiFAQ
Questions,
answered.
No independent test settles it, and taste in writing varies. Both write long drafts, follow a style you give them and work from your own files. Give both the same brief and keep the one whose draft needs fewer edits.
Both have a coding agent on paid plans, Codex in ChatGPT and Claude Code in Claude, and each company publishes coding results for its own models on its own tests. Try both on a real task from your codebase.
No. ChatGPT makes images in the app and through the API. Claude does not make photos or illustrations; it can draw diagrams and charts in code.
For one person, ChatGPT has the cheaper paid plan: Go at $6 a month, which may include ads. Plus and Claude Pro both cost $20 a month, and team seats cost the same on both. In the API, Claude's top model costs less per token, and their mid and small models cost the same.
On consumer plans ChatGPT uses your chats for training unless you opt out, and Claude asks you to choose in its privacy settings. On business plans and the API, neither trains on your data by default.
OpenAI and Anthropic each publish results for their own models, mostly on different tests run their own way, so the numbers do not line up well enough to rank the two. Each company's own model pages give its results.
