Best AI Chatbots in 2026: Free Tiers and Pricing, Verified

The best AI chatbot for most people is ChatGPT Plus at $20/month (checked August 2026): the widest toolbox plus a 256K-token reasoning context. Claude Pro, from $17/month, wins for long documents and coding. Perplexity Pro, $20/month, wins for cited research. Gemini AI Plus is the cheapest paid upgrade at $4.99/month, and free Microsoft Copilot is the best $0 answer engine. Every price and limit below comes from vendor pricing pages, help centers, and official store listings — not from other roundups.

Key takeaways

• ChatGPT Free allows unlimited GPT-5.6 Luna chats but caps Instant context at 27K tokens, roughly 12 pages; Pro extends reasoning context to 400K.
• Perplexity Free allows exactly 3 Pro searches per day and 1 Research query per month, per its own help center.
• Anthropic and Google publish paid limits only as multipliers (2x, 4x, 5x, 20x) of baselines neither vendor discloses.
• A $12.99 Microsoft 365 Family plan gives AI features to exactly one of its six members, capped at 60 credits per month.
• Cheapest paid upgrades, checked August 2026: Gemini AI Plus $4.99, ChatGPT Go $8 (App Store), Microsoft 365 Personal $9.99.

The short answer: which chatbot wins each job#

Most roundups rank chatbots by feel and recycle each other's claims. This comparison does something different. It states what each free tier concretely allows, what each paid tier concretely adds, and one documented limitation per tool, all from the vendors' own pages.

Five verified chatbots at a glance (prices checked August 2026)

Chatbot	Free tier gets you	Cheapest paid tier	Documented limitation
ChatGPT	Unlimited GPT-5.6 Luna chats; 27K-token Instant context	Go at $8/mo (App Store price)	Go may include ads
Claude	Web search, memory, code execution, connectors	Pro from $17/mo (annual billing)	No numeric usage quotas published
Gemini	3.6 Flash, Deep Research, Gemini Live, 15 GB storage	AI Plus at $4.99/mo	Paid limits are multipliers of an unpublished baseline
Perplexity	3 Pro searches/day, 1 Research query/month	Pro at $20/mo	Advanced models may throttle in heavy weeks, even on Pro
Microsoft Copilot	Web-grounded chat on Windows, macOS, iOS, Android, Edge	Microsoft 365 Personal at $9.99/mo	60 AI credits/mo; AI is owner-only on Family plans

How these numbers were verified#

Every figure in this article traces to a page fetched on August 11, 2026: the ChatGPT and Claude pricing pages, Google's Gemini subscriptions page, Perplexity's help center, Microsoft's pricing and support pages, and official App Store listings. No third-party roundup supplied any fact. Where a vendor publishes no number — and two of the five publish surprisingly few — the gap is stated outright rather than filled with a guess.

1. ChatGPT: the widest toolbox, tiered hard by context window#

What the free tier actually allows#

Free users get unlimited everyday text chats with GPT-5.6 Luna, subject to abuse-prevention safeguards. Search, projects, shared projects, study mode, the built-in browser, and using existing GPTs are all included, along with a 500 MB Library for uploads. Image generation, voice, file uploads, and deep research work too, but each carries its own limit with reset timing shown in-product.

The hard number is context: the Instant model family totals 27K tokens on Free, which OpenAI describes as roughly 12 pages of text input. Free users cannot create GPTs, schedule tasks, use voice with video, or use record mode. Consumer content is used for model training by default across all individual plans, with an opt-out available.

What each paid tier adds#

Go, at $8/month per OpenAI's App Store listing (checked August 2026), doubles Instant context to 54K tokens and lifts Reasoning context to 256K. It adds more messages with tools, more uploads and image creation, scheduled tasks, and the ability to create and share GPTs. Two caveats from OpenAI's own pages: Go does not include GPT-5.6 Sol, and the plan may include ads.

Plus, at $20/month (checked August 2026), adds GPT-5.6 Sol and advanced reasoning, expanded deep research and Codex access, record mode, developer mode in beta, and early access to new features. Pro starts at $100/month for 5x the usage of Plus; $200/month buys 20x. Both Pro tiers run a 128K Instant and 400K Reasoning context and add GPT-5.6 Sol Pro plus unlimited, faster image generation.

None of Go, Plus, or Pro offers annual billing, per OpenAI's help center. Budget accordingly: the sticker price is the monthly price.

The specific limitation#

Long documents on a budget. A 27K-token Instant context means Free stalls past roughly 12 pages, and even Plus tops out at a 256K Reasoning context against Claude's paid 1M. The cheap fix, Go, is also the tier where OpenAI says ads may appear.

2. Claude: the long-document specialist with unpublished quotas#

What the free tier actually allows#

The Free plan includes web search, memory across conversations, code generation, data visualization, file creation, code execution, extended thinking, desktop extensions, and Slack and Google Workspace connectors. Apps run on web, iOS, Android, and desktop.

The catch is capacity. The pricing page says only that usage limits apply, and the help center describes Free as limited capacity for occasional use — no message counts, no token quotas.

What each paid tier adds#

Pro costs $17/month billed annually ($200 up front) or $20 month-to-month (checked August 2026). It adds more usage, Claude Code, Research, Cowork, unlimited projects, multiple model access, and Claude for Microsoft 365. Paid plans also carry the headline spec: a 1M-token context window on Opus 5 and Sonnet 5 in chat, and 500K on Opus 4.8 through 4.6 and Sonnet 4.6.

Max starts at $100/month for 5x Pro capacity per session; $200/month buys 20x, higher output limits, priority during high-traffic periods, and early feature access. File handling is documented in detail: 500 MB per file, 20 files per chat, images up to 8000x8000 pixels, and PDFs up to 1,000 pages — though charts and images are analyzed only in PDFs of 100 pages or fewer, with longer files processed text-only.

The specific limitation#

No absolute numbers exist anywhere. Free is limited, Max is 5x or 20x Pro, and Pro itself is never quantified — there is no way to know before paying how much usage $100 buys. A second documented quirk: files added to a project cap at 30 MB each, a fraction of the 500 MB chat-upload allowance.

3. Gemini: the cheapest paid upgrade, priced in multipliers#

What the free tier actually allows#

A Google Account gets Gemini 3.6 Flash with varying access to 3.1 Pro, image generation and editing, Deep Research, Gemini Live voice conversations, Canvas, Gems, limited Nano Banana Pro in Flow, and 15 GB of cloud storage. On breadth, that free list beats some rivals' paid tiers — image generation requires a Pro subscription on Perplexity, for example.

What each paid tier adds#

AI Plus at $4.99/month is the cheapest paid chatbot upgrade among these five (checked August 2026). It doubles usage limits versus Free and adds video generation, 200 Flow credits, Gemini in Gmail and Vids, more NotebookLM Audio Overviews, and 400 GB of storage.

AI Pro at $19.99/month quadruples free limits and includes 1,000 Flow credits, higher access to Gemini 3 Pro, Deep Search, and agentic features, higher Jules coding-agent limits, 5x more Audio Overviews, Gemini in Gmail, Docs, and Vids, YouTube Premium Lite, and 5 TB of storage. AI Ultra comes in two tiers, $99.99/month and $199.99/month, reaching up to 20x higher usage than AI Pro, with 10,000 or 25,000 Flow credits, Project Genie access, a full YouTube Premium individual plan, and storage starting at 20 TB.

The specific limitation#

Every paid limit is a multiplier of a number Google never states. Doubling an unpublished free baseline is an unfalsifiable promise: there is no way to compute what any tier actually allows before subscribing.

4. Perplexity: the only vendor that publishes exact free limits#

What the free tier actually allows#

Free users get practically unlimited basic searches, exactly 3 Pro searches per day, and 1 Research query per month. There is no advanced-model access, no image generation, and no Comet Assistant browser-agent queries; file uploads are limited to basic use. Those are the vendor's own published numbers, not estimates.

What each paid tier adds#

Pro costs $20/month or $200/year per Perplexity's App Store in-app purchase listing (checked August 2026); students pay $10/month for Education Pro with SheerID verification. Pro turns on a model picker spanning GPT-5.2, Claude Sonnet 4.6, Gemini 3.1 Pro, and Grok 4, plus 10x as many citations per answer, image generation, limited video generation, and up to 50 file uploads per Space. Uploads support PDFs, CSVs, audio, video, and images.

Max costs $200/month or $2,000/year, with annual billing available only through the web app. It adds the highest tier of model access, plus higher limits than Pro on Deep Research, asset generation, and file uploads. Pro and Max subscribers can opt out of AI training on their data.

For workflows built around sourced answers, the roundup of AI research assistants covers the adjacent category in the same verified format.

The specific limitation#

Perplexity's own help center warns that advanced-model queries may be limited during weeks of especially heavy usage — a disclosure that applies to Pro, the paid tier. API access is also billed separately even at the $200 Max level.

5. Microsoft Copilot: free everywhere, credit-metered when paid#

What the free tier actually allows#

Free Copilot offers chat with real-time, web-grounded answers on Windows, macOS, iOS, Android, and in Edge — and sign-in is optional, which no other chatbot here matches. Microsoft publishes no numeric caps for this tier, so plan around features, not quotas.

What each paid tier adds#

Microsoft 365 Personal, at $9.99/month or $99.99/year (checked August 2026), puts Copilot inside desktop Word, Excel, PowerPoint, Outlook, and OneNote, includes 60 AI credits per month, and adds 1 TB of OneDrive storage. Chat and Pages allow extensive use without consuming credits; credit-metered actions include image generation, and an inbox-summary request deducts one credit. Voice runs 30 to 60 minutes daily and Vision 10 to 15 minutes daily, depending on tier.

Family, at $12.99/month or $129.99/year, covers up to six people with 6 TB of pooled storage — but the AI features belong to the subscription owner alone. Premium, at $19.99/month or $199.99/year, replaces the 60-credit cap with extensive usage for select Copilot features, adds the Researcher, Analyst, and Tasks agents, and grants preferred access to AI models during peak times.

The specific limitation#

The Family math. Six people can share the $12.99 plan, yet exactly one of them — the owner — gets Copilot in Office apps and the monthly credits; the other five get plain Chat. Microsoft's support pages state that AI benefits cannot be shared.

Free tiers compared: what $0 concretely buys#

Free-tier allowances from vendor documentation (checked August 2026)

Free tier	Documented allowance	Documented exclusions
ChatGPT Free	Unlimited Luna text chats; 27K Instant context; 500 MB Library; web search; use of GPTs	No GPT creation, scheduled tasks, voice with video, or record mode
Claude Free	Web search, memory, code execution, extended thinking, Slack and Workspace connectors	No published usage numbers; Claude Code and Research are paid features
Gemini Free	3.6 Flash, Deep Research, image generation, Gemini Live, 15 GB storage	Video generation reserved for paid tiers; Nano Banana Pro in Flow is limited
Perplexity Free	Practically unlimited basic searches; 3 Pro searches/day; 1 Research query/month	No advanced models, image generation, or Comet Assistant queries
Copilot Free	Web-grounded chat on every major platform; sign-in optional	No Office-app integration; no published caps to plan around

Two patterns stand out. Claude and Gemini give away the most features; Perplexity gives away the clearest numbers. If $0 is a hard requirement, pair one full-feature tier with Perplexity's three daily Pro searches — and see the guide to free AI tools that are actually free for the wider no-cost stack.

File and data handling: uploads and training defaults#

Claude documents the most generous chat-upload caps: 500 MB per file, 20 files per chat, and PDFs up to 1,000 pages, with visual analysis only in PDFs of 100 pages or fewer. Supported formats include PDF, DOCX, CSV, XLSX, JSON, and EPUB, with XLSX requiring code execution enabled. ChatGPT Free includes a 500 MB Library plus in-prompt file and image uploads, and paid tiers raise the unpublished upload allowances.

Perplexity accepts PDFs, CSVs, audio, video, and images, with Pro allowing up to 50 files per Space. On training defaults, OpenAI's pricing page marks every individual plan as using content for model training with an opt-out available, while Perplexity offers a training opt-out to Pro and Max subscribers and keeps enterprise data out of training entirely. Before uploading client files anywhere, run the tool through the AI tool security checklist.

Who each chatbot is not for#

Every recommendation above has a counter-audience, drawn from the same vendor pages.

• ChatGPT is not for long-document work on a free budget: a 27K-token Instant context is roughly 12 pages.
• Claude is not for anyone who needs quantified limits before subscribing, or a paid tier under $17.
• Gemini is not for buyers who want to know exact allowances: every paid limit multiplies an undisclosed baseline.
• Perplexity is not for creative or visual work on the free tier: zero image generation and one Research query per month.
• Copilot Family is not for households expecting shared AI: five of six members get plain Chat only.

Which one deserves your money: decision rules#

The reliable rule is simple: pay for the chatbot whose free tier you keep exhausting, because a limit you actually hit is the only limit that matters. From there, decide by budget.

• Under $5: Gemini AI Plus at $4.99/month, which doubles free limits and adds video generation plus 400 GB of storage.
• Under $10: ChatGPT Go at $8 via the App Store if the 27K context blocks you and possible ads are acceptable; Microsoft 365 Personal at $9.99 if Copilot inside Word and Excel is the actual goal.
• At $20: a three-way call. Claude Pro for long documents and coding, on the strength of the paid 1M-token context; Perplexity Pro for cited, multi-model research; ChatGPT Plus for the widest toolbox of images, voice, GPTs, and deep research.
• At $100 and up: only for daily heavy use. ChatGPT Pro, Claude Max, Gemini AI Ultra, and Perplexity Max mostly sell usage multiples rather than new capability, with narrow exceptions like GPT-5.6 Sol Pro on ChatGPT Pro.

Before committing, put the shortlist through a structured evaluation process before buying: a week of real workload beats any comparison table, this one included.