Which AI chatbot is actually best right now? The answer changed a lot this autumn. OpenAI launched GPT-6 Astra in early September, Anthropic followed with Claude Opus 5.5 later that month, Google’s Gemini 4 Argon became the top Gemini model, and xAI shipped Grok 4.7. Each company also reworked its plans, so the old “they all cost $20” rule no longer holds.
In this comparison I look at the four big assistants from a normal user’s point of view: which models you get, how smart they are on independent tests, what the free and paid plans include, and which one fits which kind of person. Prices and features come from each company’s official plan pages, checked on November 5, 2026, plus the independent sources listed at the end.
- Smartest model right now: Claude Opus 5.5, which leads the Artificial Analysis Intelligence Index (58), ahead of GPT-6 Astra, Gemini 4 Argon and Claude Fable 5.1 (53 each) and Grok 4.7 (46).
- Best free plan: ChatGPT Free (unlimited text chats) or Claude Free (Sonnet, files, 1M-class context depending on model). Gemini Free is now auto-routed to lighter models.
- Best $20 plan: ChatGPT Plus for all-round features, Claude Pro for coding and long documents, Google AI Pro for Google apps.
- Best for coding: Claude (Opus 5.5 plus Claude Code), then ChatGPT with Codex.
- Best for real-time social data: Grok, with Grok Imagine image and video on SuperGrok ($30).
Latest Models From Each Company
| Company | Top model | Everyday model | Small, fast model |
|---|---|---|---|
| OpenAI (ChatGPT) | GPT-6 Astra | GPT-6.1 Sol | GPT-6 Luna |
| Anthropic (Claude) | Claude Fable 5.1 / Opus 5.5 | Claude Sonnet 5.5 | Claude Haiku 5.5 |
| Google (Gemini) | Gemini 4 Argon | Gemini 3.8 Flash | Gemini Flash-Lite |
| xAI (Grok) | Grok 4.7 | Grok 4.7 / 4.3 | Grok 4.1 Fast (API) |
A quick note on Claude: Fable 5.1 is Anthropic’s most capable widely released model for long, hard reasoning, but Opus 5.5 is the one Anthropic recommends for most work and the one that tops independent rankings. In the apps, Fable is limited to paid usage credits on Pro and up to half the weekly limit on Max and Ultra.
Independent Benchmark Scores
Company benchmark tables are hard to compare, so I prefer the Artificial Analysis Intelligence Index, which runs the same suite of tests on every model. These are the best scores for each model from its October 17, 2026 snapshot:
| Model | Intelligence Index | Setting |
|---|---|---|
| Claude Opus 5.5 | 58 | max effort |
| Claude Sonnet 5.5 | 56 | max effort |
| GPT-6 Astra | 53 | max |
| Gemini 4 Argon | 53 | high |
| Claude Fable 5.1 | 53 | max effort |
| GPT-6.1 Sol | 52 | max |
| Grok 4.7 | 46 | xhigh |
A few vendor-reported numbers fill in the picture (from a September 29, 2026 comparison by Second Talent): Claude Opus 5.5 scores 89.9% on SWE-bench Pro and 81.8% on OSWorld computer use; GPT-6 Astra scores 72.6% on OSWorld 2.0 and 62.7% on ARC-AGI-3 with the standard harness; on Humanity’s Last Exam with tools, Opus 5.5 reports 67.7% against 57.2% for GPT-6 Astra. Treat vendor numbers with care, but the pattern matches the independent index: Claude leads on coding and agent tasks, and the others are close behind on general reasoning.
The newer v5.0 index (with tools): Gemini 4 Carbon moves to the top
Artificial Analysis has since moved to version 5.0 of its index, which scores models with tools turned on. Scores are not comparable with the older snapshot above, but the order is useful. These are the best settings for each model as of October 29, 2026:
| Model | Index v5.0 (best setting) | Can you use it in an app today? |
|---|---|---|
| Gemini 4 Carbon | 59.0 | No, cyber defenders only (Fairwind Program) |
| Claude Opus 5.5 | 58.4 | Yes, Claude Pro and up |
| GPT-6 Astra | 56.9 | Yes, ChatGPT Plus and up |
| GPT-6.1 Sol | 54.5 | Yes, ChatGPT Plus and up |
| Claude Sonnet 5.5 | 54.2 | Yes, including Claude Free (with limits) |
| GPT-6.1 Terra | 51.0 | Mainly via the API |
| Gemini 4 Argon | 50.4 | Yes, Google AI Pro and up |
So the practical ranking for anything you can actually subscribe to is unchanged: Claude Opus 5.5 first, then GPT-6 Astra, with Gemini’s best public model a little further back. Gemini 4 Carbon is the one to watch. I cover what we know in my Gemini 4 Carbon benchmarks post.
Benchmarks are not everything. Day to day, what matters is whether the assistant has the tools you need, which is where the plans differ most.
Free Plans Compared
| ChatGPT Free | Claude Free | Gemini Free | Grok Free | |
|---|---|---|---|---|
| Main model | GPT-6 Luna, GPT-5 Thinking Mini | Sonnet and Haiku | “Gemini Auto” (mostly Flash-Lite, Flash or Pro for hard questions) | Grok with limits |
| Chat limit | Unlimited text chats (abuse guardrails) | Limited every 5 hours and every week | Limited | Limited |
| Files, images, voice | Limited | Files, voice, web search, Artifacts | Multimodal input, Gemini Live | Image generation, voice |
| Context | 27K (Instant) | Up to 1M depending on model | Not stated | Not stated |
| Standout | No hard cap on text chats | Projects (up to 5), memory, connectors | Reasoning controls Low to High | Limited web and X search |
Google changed Gemini’s free tier on October 9, 2026: instead of a model picker, free users get “Gemini Auto”, which sends most prompts to Flash-Lite and harder ones to Flash or Pro. Free users keep multimodal input, Gemini Live and Low to High reasoning controls, but cannot choose the model.
Paid Plans Compared
| Price (US/month) | ChatGPT | Claude | Google AI (Gemini) | Grok |
|---|---|---|---|---|
| About $5 to $10 | Go (about $8, may show ads) | n/a | AI Plus ($4.99, Flash models only from Nov 13) | SuperGrok Lite ($10) |
| About $20 to $30 | Plus ($20) | Pro ($20, or $17 yearly) | AI Pro ($19.99) | SuperGrok ($30) |
| About $40 | n/a | n/a | n/a | X Premium+ ($40) |
| About $100 | Pro 100 | Max ($100) | AI Ultra ($99.99) | SuperGrok Plus ($100) |
| About $200 and up | Pro 200 ($200), Pro 500 ($500) | Ultra ($200) | Ultra top tier ($200) | SuperGrok Heavy ($300) |
What the $20 tier gets you at each company:
- ChatGPT Plus ($20): GPT-6 Astra, GPT-6.1 Sol and the rest of the GPT-6 family, deep research, Codex, projects, custom GPTs, scheduled tasks and Sites, with a 256K reasoning context. Details in ChatGPT Free vs Go vs Plus vs Pro.
- Claude Pro ($20): Opus 5.5, Claude Code, Research, Claude in Chrome and Microsoft 365, unlimited projects, and Fable through usage credits. Details in Claude Free vs Pro vs Max vs Ultra.
- Google AI Pro ($19.99): manual access to Pro models including Gemini 4 Argon, four times the standard free usage, and extras such as YouTube Premium Lite.
- SuperGrok ($30): Grok’s flagship models, Grok Imagine image and video, and connectors. See Grok pricing.
At $100, Google AI Ultra offers 5x AI Pro usage, Claude Max gives 5x Pro usage plus $100 a month of API credits, and ChatGPT Pro 100 adds 128K Instant and 400K reasoning context plus GPT-5.6 Sol Pro.
Features That Set Each One Apart
ChatGPT: the most complete all-rounder
ChatGPT has the widest set of consumer tools: GPT-Image 2.6 image generation, GPT-Live voice, deep research, Codex for coding, ChatGPT Work on desktop, custom GPTs and Sites. If you want one app that does a bit of everything well, it is still the safest pick. It also has the most generous free text chat.
Claude: best for coding, writing and long documents
Claude leads independent tests and is the favourite of many developers thanks to Claude Code, which shares your plan’s usage. Every plan, including Free, can work with very long contexts (up to 1M tokens depending on the model), and Anthropic has added Claude Design, Slides, Docs and Dashboards. The trade-off is usage limits, which reset every five hours and every week, although verified Ultra accounts no longer have the five-hour limit.
Gemini: best inside Google’s world
Gemini is built into Search, Android and Google Workspace, and its image model family (Nano Banana) is strong and cheap. Gemini 4 Argon matches GPT-6 Astra on the Intelligence Index. Note the October changes: free users are auto-routed, and from November 13 the $4.99 AI Plus plan loses manual access to Pro models.
Grok: real-time X data and image and video generation
Grok’s unique strength is live access to posts on X, which makes it useful for news, trends and social research. SuperGrok bundles Grok Imagine image and video generation. On pure reasoning it trails the other three on the independent index, and SuperGrok is $10 more than the other $20 plans.
Feature-by-Feature Comparison
This is the table I wish I had when I started paying for these. It shows what each assistant can do on its main paid plan (ChatGPT Plus, Claude Pro, Google AI Pro, SuperGrok) as of November 2026.
| Feature | ChatGPT | Claude | Gemini | Grok |
|---|---|---|---|---|
| Image generation | Yes (GPT-Image 2.6) | No native image model | Yes (Nano Banana) | Yes (Grok Imagine) |
| Video generation | Not in the Plus feature list | No | Yes (Veo, plan limits apply) | Yes (Grok Imagine) |
| Voice conversation | Yes (GPT-Live) | Yes (voice mode) | Yes (Gemini Live) | Yes |
| Deep research | Yes | Yes (Research) | Yes (Deep Research) | Yes (DeepSearch) |
| Coding agent | Codex | Claude Code | Gemini CLI / Jules | Grok Build (API) |
| Browser agent | Built-in browser, ChatGPT Work | Claude in Chrome | Gemini in Chrome with auto browse | Limited |
| Office and email | Excel, Word, PowerPoint, Sheets extensions | Microsoft 365 and Outlook | Gmail, Docs, Sheets, Drive built in | Connectors |
| Live social data | Web search | Web search | Google Search | Web plus live X posts |
| Projects and memory | Yes | Yes (unlimited projects on Pro) | Yes (Gems, memory) | Yes |
| Max context (paid) | 256K reasoning (400K on Pro) | Up to 1M, model dependent | Up to 1M on Pro models | Up to 500K to 1M, model dependent |
| Cheapest paid plan | Go, about $8 | Pro, $17/month yearly | AI Plus, $4.99 | SuperGrok Lite, $10 |
Two things stand out. Claude is the only one without its own image generator, so if pictures matter to you it should not be your only subscription. And Gemini is the only one that lives inside the apps where most people already keep their email and documents.
Privacy: Which One Trains on Your Chats?
All four can use consumer chats to improve their models, and all four let you limit that. Here is where to look:
| Assistant | Where to turn training off | Business option without training |
|---|---|---|
| ChatGPT | Settings > Data controls > Improve the model for everyone (also use Temporary Chat) | ChatGPT Business and Enterprise |
| Claude | Settings > Privacy > Help improve Claude | Claude Team and Enterprise |
| Gemini | Gemini Apps Activity (turn off “Keep activity”) | Gemini in Google Workspace business plans |
| Grok | Settings > Data controls in Grok (and X privacy settings) | xAI business and API terms |
If you handle client or company data, use a business plan from any of them rather than a personal account. If you want nothing to leave your PC at all, a local model is the only real answer.
How to Test All Four Yourself in 20 Minutes
Every one of these has a free plan, so the best way to choose is to give all four the same tasks. Open each in its own browser tab and paste these prompts, one at a time:
- Writing: “Rewrite this email so it is friendly but firm, under 120 words:” followed by a real email you need to send.
- Reasoning: “I earn X a month, my rent is Y and I want to save Z by next December. Build me a month-by-month plan and show the maths.”
- Long document: upload a PDF you know well (a contract, a manual or a report) and ask “List every deadline and obligation in this document with page numbers.” Check the answers against the document.
- Current events: “What happened in [your field] this week? Give sources.” This shows how well each one searches.
- Your daily task: whatever you will actually use it for most, such as code, lesson plans, product descriptions or study notes.
Score each answer out of 5 for accuracy and usefulness. The winner on your own tasks matters far more than any benchmark, and the test costs nothing.
Best Chatbot for Each Use Case
| Use case | Best pick | Why |
|---|---|---|
| Coding and software agents | Claude | Top coding scores, Claude Code included from Pro |
| Long PDFs, contracts, books | Claude | Up to 1M context on every plan |
| General everyday assistant | ChatGPT | Widest feature set, unlimited free text |
| Images | ChatGPT or Gemini | GPT-Image 2.6 and Nano Banana |
| Gmail, Docs, Android users | Gemini | Deep Google integration |
| News and social media trends | Grok | Live X search |
| Students on a budget | ChatGPT Free or Go, Claude Free | Strong free tiers |
| Privacy, offline, no subscription | A local model | See my best Ollama models list |
If you are a developer choosing an API rather than an app, the ranking shifts toward price. My AI API pricing comparison covers that, including DeepSeek.
Frequently Asked Questions
Which is smarter, ChatGPT or Claude?
On the Artificial Analysis Intelligence Index (October 17, 2026), Claude Opus 5.5 scores 58 and GPT-6 Astra 53, so Claude is ahead on that independent test. In daily use both are excellent; ChatGPT has more consumer tools.
Is Gemini better than ChatGPT?
Gemini 4 Argon ties GPT-6 Astra on the index. Gemini is the better choice if you live in Google apps; ChatGPT has the broader feature set.
Is Grok as good as ChatGPT?
Grok 4.7 scores lower on independent reasoning tests (46 vs 53) but offers live X data and bundled image and video generation.
Which AI chatbot is free?
All four have free plans. ChatGPT Free has unlimited text chats; Claude Free includes Sonnet, files and projects; Gemini Free uses automatic model routing; Grok Free has limited chats and search.
Can I use more than one?
Yes, and many people do: a free plan for one and a paid plan for the other is a cheap way to get the best of both.
Which AI chatbot is best for writing?
Claude is the most popular choice for natural, less robotic writing and for editing long drafts. ChatGPT is a close second and better if you also want images for the same piece.
Which one has the biggest context window?
Claude and Gemini both reach up to 1M tokens on supported models. ChatGPT Plus gives 256K for reasoning, and Pro gives 400K.
Which is best for students?
Start with the free plans. Gemini often has student offers, ChatGPT Free has unlimited text chats, and Claude Free is strong for essays and reading long PDFs. My best AI tools for students guide lists current deals.
Which one can make images and videos?
ChatGPT and Gemini make images, Gemini and Grok make videos, and Grok Imagine does both on SuperGrok. Claude does not generate images itself.
Is the most expensive plan worth it?
Only for heavy daily use. The $100 to $200 plans mostly buy more usage and priority access, not a smarter model. Every company gives its top public model on the $20 tier (Claude Fable is the exception, which is heavily limited on Pro).
Which is best for privacy?
None of the consumer apps is fully private by default. Turn off training in settings, or use a business plan. For full privacy run a local model with Ollama; start with my Ollama install guide.
Final Thoughts
In November 2026 there is no bad choice among the big four, but there are clear strengths. Claude is the smartest on independent tests and the best for coding and long documents. ChatGPT is the most complete everyday assistant. Gemini is the natural fit for Google users, and Grok is the one to pick for live social data. If you can only pay for one, decide by the tools you use most, then try the free plans of the others before committing.
For a deeper look at one matchup, see DeepSeek vs Claude, and if you hit errors with Claude, my Claude not working fix guide explains every message.
Sources
- OpenAI: ChatGPT pricing
- Anthropic: Claude plans
- Android Headlines: Google AI Plus, Pro and Ultra changes (Oct 15, 2026)
- Engadget: Google AI Ultra now starts at $100
- Krater: Grok pricing 2026
- Artificial Analysis Intelligence Index snapshot (Oct 17, 2026)
- DataLearner: Artificial Analysis Intelligence Index v5.0 table (updated Oct 29, 2026)
- Second Talent: GPT-6 vs Claude vs Gemini (Sep 29, 2026)
