ChatGPT vs Gemini vs Claude: Which AI Assistant to Use

Everyone using AI day to day eventually asks the same question: ChatGPT, Gemini, or Claude? All three now sit near or above a billion monthly users, all three got real upgrades in the last few weeks, and all three are good enough that “which is better” is the wrong question. The better one is which is better for you - a student writing essays, a developer shipping code, a researcher digging through PDFs, or someone who just wants quick, reliable answers. This page covers what each product actually costs today, what it’s strongest at, where it falls short, and who should pick what.

One naming note before the pricing: Google rebranded its enterprise cloud platform, Vertex AI, to the “Gemini Enterprise Agent Platform” in April 2026. That rebrand is about the developer/cloud product - the thing businesses use to deploy Gemini inside their own applications. It does not touch the consumer Gemini app or the gemini.google.com pricing this page covers; if you’re comparing everyday chat assistants, ignore anything that references the “Enterprise Agent Platform” - that’s a different product for a different audience. See Google Vertex AI if you actually need that side of it.

What each one is actually best at

ChatGPT (OpenAI) has the largest user base of the three - OpenAI said it passed 1 billion weekly active users on 31 July 2026 [1] - and the broadest single-app feature set: Agent mode, Deep Research, built-in image generation, voice mode, a Codex coding agent, and the widest set of third-party plugins and connectors. If you want one app that does almost everything without switching tools, ChatGPT is that app, and its size means the most tutorials, community troubleshooting, and third-party integrations exist for it.

Claude (Anthropic) has the smallest consumer footprint but the strongest reputation among people who write or code for a living. Its instruction-following is unusually reliable for long, structured prompts, it holds up well across very long documents and codebases, and Claude Code is widely regarded as one of the best agentic coding tools available. Anthropic keeps the product surface narrow - three plans, no image generation, less ecosystem sprawl - which some people read as focus and others read as a real capability gap.

Gemini (Google) wins on native multimodality and ecosystem depth: one request can take a video, a PDF, and a question together, and if your life already runs on Gmail, Docs, Drive, and Android, Gemini shows up inside all of it rather than living in a separate tab. It also currently offers the largest context window an ordinary paying subscriber can actually use inside the chat app itself, not just the API - more on that below.

Current model lineups (September 2026)

Each company ships a family of models, not one model, and the names change often enough that it’s worth knowing the shape rather than memorizing a specific version.

• OpenAI: the GPT-5.6 family (Sol, Terra, Luna) has been the generally available default since 9 July 2026; GPT-5.6 Sol and Luna are now the previous generation. GPT-6 Astra began rolling out 3 September 2026 as OpenAI’s newest and most capable model - and, by OpenAI’s own account, the first OpenAI model to cross the “Critical” threshold for cybersecurity capability in its internal safety framework [2][3]. The rollout was messy: initial access went to a small vetted cohort, Sam Altman publicly apologized for the confusion, and in early September Astra (labeled “GPT-6 Pro” in the app) reached Pro, Business, and Enterprise plans before Plus [4][5]. On 22 September 2026 OpenAI released GPT-6 Sol and GPT-6 Luna, the mid and budget tiers of the GPT-6 family below Astra, at half the API price of their GPT-5.6 predecessors [18]. As of 25 September 2026, ChatGPT’s pricing page lists GPT-6 Astra, Sol, and Luna as included on Plus, with expanded use on Pro, and none of them on Free or Go [22]; at launch, reporting put the GPT-6 models in ChatGPT’s Work mode and Codex rather than the ordinary Chat model picker, with GPT-6 Luna reaching Free and Go only in the desktop app [23].
• Anthropic: Claude Sonnet 5 (GA 30 June 2026) is the mid-tier default, Claude Opus 5.5 (GA 22 September 2026, replacing Opus 5 as the Opus model Anthropic recommends for most work) is the flagship, and Claude Fable 5.1 (GA 1 September 2026) sits above Opus for the hardest reasoning and coding work. A sibling model, Claude Mythos 5.1, exists but is restricted to vetted cybersecurity and biosecurity researchers - you will not encounter it as a regular subscriber [6]. Anthropic says Sonnet 5.5 and Haiku 5.5 are coming “in the coming weeks”; neither was out as of 25 September 2026 [19].
• Google: Gemini 3.1 Pro is Google’s top Pro model (it superseded the earlier “Gemini 3 Pro,” which is now largely retired) but is still in preview in the API, so Gemini 3.8 Flash, released 2 September 2026, is in practice Google’s flagship generally available model and its fast workhorse tier - the fourth Flash release in under four months [7]. Gemini 3.8 Live, Google’s real-time voice model, reached general availability in the Gemini API on 15 September 2026 [20]. Deep Think is Google’s extended-reasoning mode, reserved for the top subscription tier.

What each one costs

Prices below are the current US consumer rates as of early September 2026, fetched directly from each vendor’s own pricing page where possible [8][9][10]. All three change pricing often enough that you should treat this as a snapshot, not a permanent fact - click through and check before you buy.

Two things stand out. First, Claude has no cheap paid tier - you go from free straight to $17-20/month, while ChatGPT and Gemini both offer a sub-$10 middle option. Second, Gemini is the only one of the three with no mid-priced path to its best reasoning mode: Deep Think is reserved entirely for AI Ultra ($99.99+/month), while Anthropic bundles Opus into the $20/month Claude Pro plan and OpenAI puts GPT-5.6 Sol and now GPT-6 Astra on the $20/month Plus plan. (All three vendors’ top all-around tiers - AI Ultra, Claude Max, ChatGPT Pro - land in the same $100-$200 band; the gap is in what the cheaper mid-tier plans include, not in what the priciest tier costs.)

Context window: the number that’s more confusing than it looks

“Context window” gets thrown around as a single spec, but what a paying subscriber actually gets inside the chat app is not the same as what the underlying API supports, and the gap differs by vendor in a way worth knowing before you assume a bigger number automatically applies to you.

Claude’s consumer context window is now listed as “up to 1M, varies by model.” When this page was first written (4 September 2026) it reported a 200,000-token ceiling in the consumer app on every plan, while Sonnet 5, Opus 5.5, and Fable 5.1 support 1 million tokens through the API and in Claude Code [11]. As of 25 September 2026, Anthropic’s own pricing page lists the context window for Free, Pro, and both Max tiers as “Up to 1M, varies by model” [21]. That wording does not say which model gets which window on which plan, so if long-document work inside claude.ai is the reason you’re paying, test it with your own files rather than assuming the full million.

Gemini goes the other direction: AI Pro and AI Ultra subscribers get the full 1-million-token window inside the ordinary Gemini app, confirmed directly on Google’s own pricing page [9]. That makes Gemini, at $19.99/month, the one of the three that most plainly documents a full 1M window as a native part of the paid consumer product.

ChatGPT’s consumer context window is much smaller than its API’s. OpenAI’s API documents a 1.05-million-token window for GPT-5.6 and GPT-6 [12][18], but the window inside the ChatGPT app varies by mode and plan. As of 25 September 2026, ChatGPT’s pricing page lists the Instant-mode window at 27K tokens on Free, 54K on Go and Plus, and 128K on Pro, and the reasoning-mode window at 256K on Go and Plus and 400K on Pro (“varies” on Free) [22]. If long-document work inside the chat app itself is your actual use case, test it with your own documents rather than trusting either vendor’s headline figure.

Real limitations, not the marketing version

ChatGPT’s biggest practical downside is its own size and sprawl. Six subscription tiers (Free, Go, Plus, Pro at two price points, Business, Enterprise) is harder to navigate than Claude’s three or Gemini’s four, and GPT-6 Astra’s rollout - gated access, a CEO apology over the confusion, and a Pro-before-Plus staged release - followed by GPT-6 models that appeared in the separate Work mode rather than the Chat picker, shows OpenAI’s newest capability isn’t always cleanly available where you expect it [4][5][23]. Free-tier images, voice, and file uploads stay capped even though free-tier text messaging became effectively unlimited in August 2026.

Claude’s biggest practical downside is the narrowest feature set of the three. There is still no native image generation - Anthropic has no first-party text-to-image model, so “draw me a picture” sends a Claude user elsewhere every time. Usage limits on Free and Pro aren’t published as a fixed message count either; Anthropic describes them relatively (“at least 5x more than Free”) rather than in absolute terms, which makes it hard to know in advance what a given plan gets you in a busy week [13].

Gemini’s biggest practical downside is version churn and a paywalled ceiling on its best reasoning. Four Flash releases in under four months (3.5 → 3.6 → 3.7 → 3.8) between May and September 2026 is genuinely hard to track, and the free and even AI Plus tiers run a version behind whatever Google just shipped [7][9]. Deep Think, Gemini’s strongest reasoning mode, is reserved for AI Ultra - so the best version of Gemini costs $99.99 a month minimum, well above where Claude Opus or OpenAI’s Sol-tier model sit on a $20 plan.

Which one fits which kind of user

Casual, everyday user. Start free on whichever ecosystem you’re already in - Gemini if your life runs on Gmail and Android, ChatGPT otherwise, since it has the largest community of tutorials and troubleshooting if you get stuck. Don’t pay for anything until a free tier’s limits actually bother you.

Student. Check the free-tier deals before paying for anything: Google is giving US college students a full free year of AI Pro (international students get a free year of AI Plus) through 31 December 2026 sign-up [14], and OpenAI is running a Back to School promotion giving eligible US students four free months of ChatGPT Plus through 31 October 2026 [15]. Anthropic offers no individual student discount at all - Claude access for students depends on your university having a campus-wide Claude for Education agreement [16]. If your school doesn’t have one, Claude is the most expensive of the three for a student to reach.

Writer. Claude, for most people who care about prose quality - its writing tends to sound less generically “AI” and holds tone and style instructions more reliably across a long piece. ChatGPT’s Canvas mode is a reasonable second choice if you also want inline image generation for the same piece of work.

Developer. Claude Code remains the most-cited agentic coding tool of the three, but ChatGPT’s Codex and Agent mode are a legitimate alternative, especially if you’re already paying for Plus for other reasons. Gemini’s coding tooling (the newer Antigravity CLI, replacing Gemini CLI) is the least mature of the three as of this writing.

Researcher. Gemini’s combination of Deep Research, native PDF/video ingestion, and the full 1M-token context window inside the app at $19.99/month is hard to beat for source-heavy work. Claude is the stronger choice specifically for synthesizing many long documents into careful, well-cited prose once you’ve gathered the sources.

Nobody needs all three. If you’re paying for a subscription rather than staying free, pick based on the single task you do most - writing, coding, research, or everything-in-one-app - rather than trying to average across all of them.

What this page can’t tell you

Exact usage limits on any plan (none of the three publish a fixed message count), how GPT-6 Astra, Sol, and Luna map onto ChatGPT plans once the rollouts settle, and whether a given promotional student offer is still live by the time you read this - check the linked official pages directly, since all three vendors change these terms with little notice.

Further reading

• AI subscription pricing 2026 : the broader pricing comparison, including Microsoft 365 Copilot and Perplexity alongside these three.
• The 2026 LLM landscape : every other model family, for context on where these three sit in the wider market.
• Claude vs ChatGPT : the deeper enterprise/API-side comparison, for readers building on these platforms rather than chatting with them.
• What is ChatGPT , What is an LLM : background if any of this is new territory.
• Google Gemini , Claude by Anthropic : the fuller model-lineup and pricing detail behind the tables above.
• Google Vertex AI : the enterprise-cloud rebrand (“Gemini Enterprise Agent Platform”) this page distinguishes from the consumer Gemini app.
• Context window (glossary) : what a context window actually is and why bigger isn’t always usable.
• Astra becomes the first OpenAI model to cross the “Critical” cyber threshold , Google ships Gemini 3.8 Flash , GPT-6 Sol and GPT-6 Luna , and Claude Opus 5.5 : this wiki’s coverage of the releases referenced above.

Sources

1. OpenAI, confirming more than 1 billion weekly active users (31 July 2026), widely reported including by Axios and other outlets tracking OpenAI usage disclosures.
2. OpenAI, “GPT-6 Astra: A new generation of intelligence” (3 September 2026): https://openai.com/index/gpt-6-astra/
3. CNBC, “OpenAI announces rollout of GPT-6 Astra model” (3 September 2026): https://www.cnbc.com/2026/09/03/open-ai-astra-gpt-6-cyber.html
4. Dataconomy, “OpenAI Launches GPT-6 Astra With Limited Rollout Starting Today” (4 September 2026): https://dataconomy.com/2026/09/04/gpt-6-astra-launch-openai-limited-rollout/
5. Simon Willison, “GPT-6 Astra” (3 September 2026), on rollout scope, pricing, and benchmark detail: https://simonwillison.net/2026/Sep/3/gpt6-astra/
6. Anthropic, “Introducing Claude Fable 5.1 and Claude Mythos 5.1” (1 September 2026) - see this wiki’s Fable 5.1 reaches GA, Mythos 5.1 stays gated
7. Google, “Introducing Gemini 3.8 Flash and 3.8 Flash Cyber” (2 September 2026) - see this wiki’s Gemini 3.8 Flash coverage
8. Anthropic, Claude pricing page, fetched 4 September 2026: https://claude.com/pricing
9. Google, Gemini subscriptions pricing page, fetched 4 September 2026: https://gemini.google/subscriptions/
10. Aggregated from multiple current trackers cross-checked against official OpenAI Help Center pricing detail, since chatgpt.com/pricing blocks automated fetching: https://www.cloudzero.com/blog/how-much-does-chatgpt-cost/ , OpenAI Help Center, “About ChatGPT Pro tiers”: https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers
11. Claude.com pricing page and Anthropic model documentation, on the 200K consumer-app ceiling versus the 1M API/Claude Code window, fetched 4 September 2026: https://claude.com/pricing , https://platform.claude.com/docs/en/models/overview
12. OpenAI Developer Platform, pricing and models documentation (GPT-5.6 1.05M context window), fetched 4 September 2026: https://developers.openai.com/api/docs/pricing
13. Anthropic Support, “What is the Pro plan?”: https://support.claude.com/en/articles/8325606-what-is-the-pro-plan
14. Google, “College students get 12 months of Google AI free” (August 2026): https://blog.google/innovation-and-ai/products/gemini-app/student-offer-google-ai/
15. OpenAI Help Center, “ChatGPT Back to School offer for students”: https://help.openai.com/en/articles/20001493-chatgpt-back-to-school-offer-for-students
16. Anthropic, on Claude for Education as an institution-level product rather than an individual student discount, cross-checked across multiple current pricing trackers, e.g. https://costbench.com/software/ai-chatbots/claude/discounts/
17. Google, “Google Vertex AI is now the Gemini Enterprise Agent Platform” (22 April 2026) - see this wiki’s Google Vertex AI for the full rebrand detail and why it doesn’t touch the consumer Gemini app covered on this page.
18. OpenAI Developer Platform, API changelog (22 September 2026: GPT-6 Sol and GPT-6 Luna released) and pricing documentation, fetched 25 September 2026: https://developers.openai.com/api/docs/changelog , https://developers.openai.com/api/docs/pricing
19. Anthropic, “Claude Opus 5.5” (22 September 2026): https://www.anthropic.com/claude-opus-5-5 ; Anthropic models overview, fetched 25 September 2026: https://platform.claude.com/docs/en/about-claude/models/overview
20. Google, “Gemini 3.8 Live with Live Avatar” and Gemini API changelog (15 September 2026 GA): https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-live-with-live-avatar/ , https://ai.google.dev/gemini-api/docs/changelog
21. Anthropic, Claude pricing page, plan comparison (“Context window: Up to 1M, varies by model” on Free, Pro, Max 5x, and Max 20x), fetched 25 September 2026: https://claude.com/pricing
22. OpenAI, ChatGPT pricing page - individual plans and “Compare features across plans” (GPT-6 model access by plan, Instant and reasoning context windows by plan), fetched 25 September 2026: https://chatgpt.com/pricing
23. TechCrunch, “OpenAI launches GPT-6 Sol and Luna, boasting lower cost and fewer mistakes” (22 September 2026): https://techcrunch.com/2026/09/22/openai-launches-gpt-6-sol-and-luna/ ; TechRadar, “ChatGPT got GPT-6, but you can’t use it in Chat” (23 September 2026): https://www.techradar.com/ai-platforms-assistants/chatgpt/chatgpt-got-gpt-6-but-you-cant-use-it-in-chat-confused-its-time-we-talked-about-the-difference-between-chat-and-work