ChatGPT vs Claude vs Gemini: Pricing and Plans (August 2026)

Pick Claude if you work with long documents or want a serious coding tool included at $20 — Claude Pro bundles Claude Code and a 200K context window. Pick ChatGPT if you want the widest toolset: voice, image generation, custom GPTs, scheduled tasks, and study mode. Pick Gemini if you live inside Google's apps or want the cheapest paid plan of the three — Google AI Plus costs $4.99/month (checked August 2026). The privacy trade-offs differ more than the prices do.

Key takeaways

• All three mid tiers cluster near $20: ChatGPT Plus at $20, Claude Pro at $20 ($17/month billed annually), Google AI Pro at $19.99 (checked August 2026).
• Google AI Plus at $4.99/month is the cheapest paid tier anywhere in this comparison; Claude sells nothing between $20 and $100.
• Context windows depend on the plan you buy: ChatGPT's instant model gets 27K tokens on Free and 54K on Plus, Claude's consumer plans get 200K, and Gemini's AI Pro tier advertises 1M.
• ChatGPT and Gemini use consumer chats for training by default with opt-outs; Claude trains on chats only if you enable model improvement, a chat is safety-flagged, or you join a tester program.
• Gemini chats that human reviewers have seen are retained for up to three years even after you delete your activity.

Quick verdict: pick by decision rule#

Choose Claude when your work is long and text-heavy. Every Claude consumer plan carries a 200K context window, and the $20 Pro plan includes Claude Code, unlimited projects, and Research. That combination suits editors, analysts, and developers who feed the model entire codebases or manuscripts.

Choose ChatGPT when you want one subscription that does the most things. The consumer lineup spans voice, image generation, deep research, Codex, projects on every plan, and study mode on every plan. The trade-off is a small context window on the instant model unless you pay for Pro.

Choose Gemini when you already pay for Google storage or want to spend less than $5. Google AI Plus at $4.99/month doubles free-tier usage and adds video generation plus 400 GB of storage. The $19.99 AI Pro tier is the only plan in this comparison that advertises a 1M-token context window.

What each plan costs (checked August 2026)#

The flat "$20 each" framing you see in most comparisons is out of date. ChatGPT and Google now run a four-rung ladder — free, a budget tier, a $20-class tier, and a $100-class tier with a 5x-or-20x usage selector — while Claude jumps straight from free to $20 to $100.

Consumer plan pricing, checked August 2026

Plan	Price	Notable inclusions
ChatGPT Free	$0	Unlimited text chat with GPT-5.6 Luna (abuse guardrails apply); limited images, deep research, Codex, memory
ChatGPT Go	Low-cost tier; priced regionally, and OpenAI says it may adjust pricing over time	More messages, uploads, and images; longer memory; may include ads
ChatGPT Plus	$20/month	Priority access, higher limits, voice, image generation, file uploads, custom GPTs
ChatGPT Pro	From $100/month (5x or 20x usage)	GPT-5.6 Sol Pro reasoning; maximum deep research and Codex; faster image creation
Claude Free	$0	Chat on web, iOS, Android, desktop; code generation, web search, memory, file creation, extended thinking
Claude Pro	$20/month ($17/month billed annually)	Claude Code, unlimited projects, Research
Claude Max	From $100/month (5x or 20x Pro usage)	Higher output limits, priority access at high-traffic times
Gemini Free	$0	Gemini 3.6 Flash, limited Gemini 3.1 Pro, Deep Research, Gemini Live, Canvas, Gems, 15 GB storage
Google AI Plus	$4.99/month	2x usage limits, video generation, 200 Flow credits, 400 GB storage
Google AI Pro	$19.99/month	4x limits, 1M context, Jules coding agent, entry-level Antigravity, 5 TB storage, YouTube Premium Lite
Google AI Ultra	$99.99/month (5x) or $199.99/month (20x)	Deep Think, Project Genie, 10,000 or 25,000 Flow credits, from 20 TB storage

Two footnotes matter. Claude Pro's annual price is $200 billed up front, which works out to $17/month. ChatGPT Go's page carries the line "This plan may include ads," and OpenAI's help center confirms it may start testing ads on that tier.

The top of every ladder looks the same. ChatGPT Pro and Claude Max both start at $100/month with a choice of 5x or 20x usage, and Google AI Ultra publishes $99.99 for 5x and $199.99 for 20x. If your usage might climb that high, run the numbers against API pricing first — the LLM API cost estimation walkthrough shows how to compare a flat subscription against metered tokens.

The budget tiers most comparisons skip#

Google AI Plus is the quiet bargain here. For $4.99/month you get double the free tier's usage limits, video generation, 200 Google Flow credits, and 400 GB of storage that also covers Drive, Gmail, and Photos. If a free tier keeps cutting you off but $20 feels steep, this is the obvious first upgrade.

ChatGPT Go sits in the same bracket but without a settled price. It layers more messages, uploads, and image generation on top of Free, plus longer memory, and it is sold in every country where ChatGPT operates. The catch: pricing varies by region, OpenAI says it may adjust Go pricing over time as it gathers feedback, and the plan may include ads.

Claude has no answer at this level. The jump from Free goes straight to $20 Pro, and the jump after that goes straight to $100 Max. Budget-conscious users comparing entry prices should read the roundup of genuinely free AI tools before committing to any subscription.

Context windows by plan, not by press release#

Headline context numbers usually describe the best model on the best plan. The pricing pages tell a more useful story: your window depends on what you pay.

Published context windows by consumer plan, August 2026

Plan	Instant model (total context)	Reasoning model (total context)
ChatGPT Free	27K tokens	Varies
ChatGPT Go	54K tokens	256K tokens
ChatGPT Plus	54K tokens	256K tokens
ChatGPT Pro	128K tokens	400K tokens
Claude Free, Pro, and Max	200K tokens across consumer plans	200K tokens across consumer plans
Google AI Pro	1M tokens (the only Google tier with a published figure)	1M tokens (the only Google tier with a published figure)

Read the ChatGPT rows carefully. A free user pasting a long contract into the default instant model works inside 27K total tokens, and OpenAI's own table caps instant-model input on Free at roughly 12 pages of text — less than the total window. Paying $20 doubles the instant window to 54K and raises the reasoning window to 256K.

Claude flattens this entirely: 200K on every consumer plan, free included. Gemini publishes 1M on the AI Pro tier and lists no figure for Free, Plus, or Ultra on its U.S. plans page. If long inputs are your daily reality, that single spec should weigh more than any feature list.

Best for students#

Start free, and start with two accounts rather than one. Gemini's free tier includes Deep Research, Gemini Live, Canvas, and Gems, while ChatGPT Free offers unlimited everyday text chat plus study mode, which OpenAI ships on every plan including Free. Claude Free adds web search, file creation, and extended thinking at no cost.

When free limits pinch, the decision rule is simple: pay $4.99 for Google AI Plus before paying $20 for anything. Doubled limits, video generation, and 400 GB of storage cover most coursework loads. Move to a $20 tier only if a specific feature — Claude's projects, ChatGPT's custom GPTs — becomes essential.

Best for professional writers#

Claude Pro is the writing pick, and the reason is structural rather than stylistic. Long-form editing lives or dies on how much text the model can hold at once, and 200K tokens beats ChatGPT's 54K instant window at the same price. Projects give each book or client a persistent workspace.

ChatGPT Plus wins a narrower case: writers who produce illustrated content. Image generation, voice drafting, and custom GPTs ship in the $20 plan, and the reasoning model's 256K window actually exceeds Claude's for deliberate, slower work. If your deliverables mix text and visuals, Plus is the better single subscription.

Gemini suits writers embedded in Google Docs and Drive. Canvas and Deep Research are available from the free tier, and the AI Pro tier's 1M-token window is unmatched for dumping an entire research corpus into one session. Test any of the three against your own material using a repeatable process like the one in the AI output quality testing guide.

Best for developers#

Claude Pro takes this category for most working developers. Claude Code — a terminal-based agentic coding tool — is included in the $20 plan rather than sold separately, and the 200K window comfortably holds multi-file changes with their surrounding context. Nothing extra to buy, nothing to configure beyond login.

ChatGPT's coding story scales with spend. Codex access is limited on Free and Go, expands on Plus, and reaches maximum task volume on Pro from $100/month. That makes ChatGPT the pick for teams already deep in OpenAI tooling who will genuinely use Pro-level Codex throughput.

Google AI Pro is the wildcard: Jules and Antigravity access at $19.99 undercuts everyone on agent tooling per dollar. For a deeper head-to-head on coding assistants specifically, see the AI coding assistants comparison.

What each company does with your chats#

Pricing pages get all the attention; privacy pages decide who should actually get your data. These are the official policies as published by each vendor, checked August 2026.

Consumer data handling per official policy pages, August 2026

Policy question	ChatGPT	Claude	Gemini
Trains on your chats?	Yes, by default on every consumer plan; opt-out available	Only if you enable model improvement, a chat is safety-flagged, or you join tester programs	Yes, when Keep Activity is on — and it is on by default
Opt-out mechanism	Toggle named 'Improve the model for everyone'; applies account-wide, changeable anytime	Model-improvement setting in privacy settings	Turn Keep Activity off
Private chat mode	Temporary Chats: never used for training, deleted after 30 days	Incognito chats: never used to improve Claude, even with model improvement enabled	Temporary chats: not used to train Google's AI models, kept for 72 hours
Human review	Temporary Chats may be reviewed for abuse monitoring	Safety-flagged chats can be reviewed	A subset of chats is reviewed by humans, including trained service providers
Retention wrinkle	Temporary Chats purge after 30 days	Deleted chats leave back-end storage within 30 days; with model improvement on, de-identified data may persist in training pipelines up to 5 years	Human-reviewed chats are kept up to 3 years even after you delete your activity; auto-delete defaults to 18 months

The pattern is stark. ChatGPT and Gemini both start with training enabled and put the burden on you to opt out. Claude's policy is conditional: consumer chats feed training only if you enable model improvement, a chat is safety-flagged, or you join a tester program — and Incognito chats stay out regardless.

Deletion means different things at each company. Anthropic removes deleted conversations from chat history immediately and from back-end storage within 30 days, though chats already inside an in-progress training run stay in that run. Google's auto-delete defaults to 18 months, and any chat a human reviewer has handled survives your deletion for up to three years.

If chats will ever contain client material, treat these settings as part of onboarding, not an afterthought. The AI tool security checklist covers what to verify before pasting anything sensitive into any chatbot.

The honest limitation of each#

ChatGPT: default training and a 27K free window#

Every ChatGPT consumer plan trains on your content unless you flip the opt-out, and the free tier's instant model works inside just 27K total tokens — the smallest published window in this comparison. Ads may also arrive on the Go tier. Power and breadth cost you vigilance.

Claude: nothing between $20 and $100#

Claude's ladder has a missing rung. Outgrow Pro's usage limits and the only upgrade is Max, which starts at five times the price. There is no $40 or $50 middle tier, so heavy-but-not-professional users face an awkward jump that ChatGPT Go and Google AI Plus users never encounter at the bottom of their ladders.

Gemini: human review that outlives deletion#

Gemini's default settings route a subset of chats to human reviewers, including third-party service providers, and reviewed chats are retained up to three years even after you delete your activity. Turning Keep Activity off still leaves a 72-hour retention window. The $4.99 price is real, but so is this trade.

Who this comparison is not for#

These are consumer subscriptions, and three groups should look elsewhere. Developers building products need API pricing, which is billed separately — OpenAI's help center states outright that API usage is independent of ChatGPT Plus. Metered tokens follow completely different math than flat monthly plans.

Teams needing admin controls, centralized billing, or contractual data terms belong on the business tiers, which all three vendors sell separately from the plans covered here. Consumer privacy toggles are not a substitute for an enterprise agreement.

Anyone handling regulated or confidential material under strict rules should not rely on default consumer settings at all. Before buying any plan, run it through the framework for evaluating AI tools before paying.

Bottom line#

Default to Claude Pro at $20 if long documents, code, or training that happens only when you enable it drive your choice — accepting that outgrowing it means a jump to $100. Take ChatGPT Plus if breadth wins: images, voice, custom GPTs, and a 256K reasoning window in one bill, once you have switched off the training toggle.

Take Google AI Pro if the 1M-token window or the Google ecosystem matters, and Google AI Plus at $4.99 if price matters most — after reading the retention policy with clear eyes. Whichever you pick, revisit the pricing page before renewal; every figure here was checked August 2026, and these ladders keep moving.