- The main paid tiers cost about the same as of October 20, 2026: Claude Pro $20 a month ($17 billed annually), ChatGPT Plus $20 and Google AI Pro $19.99. Only Google and OpenAI sell cheaper plans: Google AI Plus at $4.99 and ChatGPT Go at $8.
- Context differs most at the $20 level. Claude's paid plans and Google AI Pro list a 1M-token window for their current models, while ChatGPT Plus lists 54K tokens for Instant and 256K for reasoning, per each company's help or pricing pages.
- OpenAI's help center says ads appear on ChatGPT Free and Go; Anthropic's and Google's plan pages mention no ads. Independent evaluators disagree: LMArena's October 15 text snapshot puts Gemini 4 Argon first, and Artificial Analysis's Intelligence Index v4.3.2 scores Claude Opus 5.5 at 58 against 53 for Argon and GPT-6 Astra.
Contents
Claude vs ChatGPT vs Gemini is a close call in October 2026. The three main consumer tiers cost about $20 a month, and each now runs a new model generation. Anthropic's Claude plans use Opus 5.5, Sonnet 5.5 and Haiku 5.5. ChatGPT moved to GPT-6 Sol and GPT-6 Luna in its Chat tab on October 7–8. Google put Gemini 4 Argon in the Gemini app for paying subscribers on October 13.
This comparison is based on Anthropic's, OpenAI's and Google's pricing pages, help centers and release notes, plus results published by LMArena, Artificial Analysis and Vals AI, all checked October 20, 2026. Prices are US list prices before tax. For deeper two-way comparisons, see ChatGPT vs Claude for coding and writing and ChatGPT vs Gemini for research.
The short version
- Price: the main tiers are Claude Pro ($20, or $17 a month billed annually), ChatGPT Plus ($20) and Google AI Pro ($19.99). Only Google and OpenAI sell cheaper plans: Google AI Plus at $4.99 and ChatGPT Go at $8.
- Context: Claude's paid plans and Google AI Pro list 1M tokens. ChatGPT Plus lists 54K (Instant) and 256K (reasoning). ChatGPT's largest window, 400K, comes with Pro, which starts at $100.
- Top tiers: Claude Max $100 or $200, ChatGPT Pro $100, $200 or $500, Google AI Ultra $99.99 or $199.99.
- Ads: of the three, only OpenAI documents ads, on ChatGPT Free and Go.
- Coding: Claude Code from Pro up, Codex on every ChatGPT plan (limited on Free and Go), and Google Antigravity with Gemini 4 Argon on Ultra only.
- Our read: no plan wins on everything. The table at the end maps jobs to plans.
How much do Claude, ChatGPT and Gemini cost?
| Price point | Claude | ChatGPT | Gemini (Google AI plans) |
|---|---|---|---|
| Free | Free: Sonnet and Haiku | Free: GPT-6 Luna in Chat; may show ads | No plan: Flash-Lite named on Google's plan page |
| Under $10 | — | Go, $8: Luna with higher limits; may show ads | AI Plus, $4.99: Flash, 2x free usage |
| About $20 | Pro, $20 ($17/month billed annually, $200 up front) | Plus, $20 | AI Pro, $19.99 |
| About $100 | Max 5x, $100 | Pro 100, $100 | AI Ultra, $99.99 (5x Pro usage) |
| About $200 | Max 20x, $200 | Pro 200, $200 | AI Ultra, $199.99 (20x Pro usage) |
| $500 | — | Pro 500, $500 | — |
| Team seat | Team Standard $25 monthly / $20 annual; Premium $125 / $100; 2–150 seats | Business Standard $25 / $20; Premium $125 / $100; 2–200 users | Google Workspace from $7 per user a month (Business Starter) |
| Enterprise | $20 per seat (annual, 20-seat minimum) plus usage at API rates | Contact sales | Contact sales |
Table 1The main tiers sit within a cent of $20; only Google and OpenAI sell plans under $10, and only ChatGPT has a $500 tier. Source: each company's pricing pages, checked October 20, 2026.
Last verified: October 20, 2026 (claude.com/pricing and the Max plan help article; learn.chatgpt.com pricing, OpenAI's Pro tiers article and Business pricing page; gemini.google/subscriptions and workspace.google.com/pricing; full rows with sources in P22-plans-compared.csv)
Claude Max and ChatGPT Pro are billed monthly only. OpenAI cut the allowance for new Pro 200 subscribers on September 29; earlier subscribers keep the old one through October 29, 2026. Claude's Max and Team plans include monthly API credits ($100 on Max 5x, $200 on Max 20x, up to $500 pooled on Team) for the Claude Platform, not the apps. For every ChatGPT tier in detail, see ChatGPT plans compared.
Which models does each plan get?
| Plan level | Claude | ChatGPT | Gemini |
|---|---|---|---|
| Free | Sonnet and Haiku; no Opus or Fable | Chat: GPT-6 Luna. Work and Codex: Luna, limited | Flash-Lite (plan page) |
| Cheap tier | — | Go: same models as Free, higher limits | AI Plus: Flash |
| About $20 | Opus 5.5, Sonnet 5.5, Haiku 5.5; Fable 5.1 only through pay-as-you-go usage credits | Chat: GPT-6 Sol. Work and Codex: GPT-6.1 Sol, GPT-6 Luna and limited GPT-6 Astra | Gemini 4 Argon, plus Pro and Flash models |
| Top tier | Max: the same models, and up to 50% of weekly limits on Fable models | Pro: Astra across the full allowance; the "Pro" reasoning option in Chat runs on Astra | Ultra: "highest access" to Argon |
Table 2Claude Opus 5.5, GPT-6 Sol in Chat and Gemini 4 Argon all start at the $20 tier; the free plans get Sonnet and Haiku, GPT-6 Luna and Flash-Lite. Source: each company's plan pages, help articles and release notes, checked October 20, 2026.
Last verified: October 20, 2026 (claude.com/pricing and Anthropic's "Claude Fable models on your plan" article; OpenAI's October 7 release notes and pricing pages; Google's subscriptions page and October 13 Argon launch post)
ChatGPT runs different models in different places. Since October 7–8, the Chat tab uses GPT-6 Sol on paid plans and GPT-6 Luna on Free and Go, per OpenAI's release notes. GPT-6.1 Sol "is available in Work and Codex, not Chat."
Claude uses one model list across chat, Claude Code and Cowork. The exception is Fable, Anthropic's most expensive model. On Pro and Team Standard seats, Fable 5 and 5.1 aren't part of plan limits and are billed to usage credits from the start. Max and Team Premium seats can spend up to half their weekly limits on Fable, according to Anthropic's help center. Our explainer on why Fable can look unavailable covers the history.
Gemini 4 Argon needs AI Pro or Ultra. Google's October 13 launch post names "Google AI Ultra and Pro subscribers" for the Gemini app and AI Mode in Search. Google's plan page agrees: its "Includes Gemini 4 Argon" badge sits on the AI Pro and Ultra cards, and AI Plus's feature list names only Flash. Our Gemini 4 Argon how-to covers each surface.
How much can each plan read at once?
| Plan level | Claude | ChatGPT | Gemini |
|---|---|---|---|
| Free | Not published for Free | 27K tokens (Instant); reasoning "varies" | 32K tokens |
| Cheap tier | — | Go: 54K (Instant), 256K (reasoning) | AI Plus: 128K |
| About $20 | 1M tokens in chat for Opus 5.5, Sonnet 5.5, Haiku 5.5 and Fable 5.1 | Plus: 54K (Instant), 256K (reasoning) | AI Pro: 1M |
| Top tier | 1M (same models) | Pro: 128K (Instant), 400K (reasoning) | Ultra: 1M |
| File size per upload | 500 MB; up to 20 files per chat; PDFs up to 1,000 pages | 512 MB; 2M tokens per text file; Free: 3 uploads a day | 100 MB (video 2 GB); up to 10 files per prompt |
Table 3Claude's paid plans and Google AI Pro and Ultra list 1M tokens; ChatGPT's largest window is 400K, on Pro. Source: each company's help pages, checked October 20, 2026.
Last verified: October 20, 2026 (Claude Help Center context-window and upload articles; chatgpt.com/pricing and OpenAI's file-uploads FAQ; Google's Gemini Apps limits and upload help pages)
These are app figures, not API limits. When a Claude chat nears its limit with code execution on, Claude summarizes earlier messages. OpenAI notes that system instructions and memory take part of ChatGPT's window. A big window also isn't the same as good recall deep inside a long document.
How do usage limits work?
None of the three publishes fixed message counts for its chat apps. Here is what each says.
- Claude: paid plans have a session limit that resets every five hours plus a weekly limit. Claude chat, Claude Code and Claude Desktop draw on the same allowance. Pro gives "at least 5x" Free usage per session, and Max gives 5x or 20x Pro. With usage credits turned on, work continues on credits instead of stopping.
- ChatGPT: Free and Go get "unlimited text chats subject to abuse guardrails," per OpenAI's pricing page, with separate limits on uploads, images and data analysis. For Work and Codex, OpenAI estimates 5–45 local messages per five hours on GPT-6 Astra, 15–160 on GPT-6.1 Sol and 350–3,000 on GPT-6 Luna for Plus and Business Standard. It says these "are not fixed message limits." Pro "currently" has no five-hour limit.
- Gemini: limits refresh every five hours until you reach a weekly limit. AI Plus gets 2x the free allowance, AI Pro 4x, and Ultra 5x or 20x AI Pro. When subscribers hit the limit, they can continue on Flash-Lite. Deep Research costs more usage.
Last verified: October 20, 2026
Ads, privacy and training on your chats
| Claude (Free, Pro, Max) | ChatGPT (Free, Go, Plus, Pro) | Gemini (personal accounts) | |
|---|---|---|---|
| Ads | None mentioned on Anthropic's plan pages | Free and Go only; Free can switch to an "Ads-Free" mode with lower limits | None mentioned on Google's plan pages |
| Training on chats | Only if you turn on Model Improvement | Can be used unless you turn off Improve the model for everyone | Yes while Keep Activity is on |
| Retention | Up to 5 years de-identified if you opt in; 30 days otherwise (Anthropic, August 28, 2025) | Temporary Chat kept up to 30 days | Auto-delete after 18 months by default (3 or 36 months, or off) |
| Private mode | Incognito chats, never used for training | Temporary Chat | Temporary chats, kept 72 hours |
| Human review | Flagged chats may be kept up to 2 years | Not covered on the data-controls page | A subset of chats; reviewed chats kept up to 3 years |
| Team and business plans | Not trained on by default (Commercial Terms) | Business, Enterprise and Edu not trained on by default | Workspace terms apply |
Table 4Only OpenAI documents ads, and only Anthropic makes training on consumer chats opt-in. Source: Anthropic's, OpenAI's and Google's privacy and help pages, checked October 20, 2026.
Last verified: October 20, 2026 (Anthropic privacy center and consumer-terms post; OpenAI's "Ads in ChatGPT" and data-controls articles; Google's Gemini Apps Privacy Hub)
OpenAI's help center says ads appear below answers, and that it is testing a multi-image ad format shown while an image is generating. Ads don't appear in Temporary Chats, next to sensitive topics or for accounts identified as under 18. See ads in ChatGPT explained for the controls.
Which coding tool comes with each plan?
| Claude Code | Codex | Google Antigravity | |
|---|---|---|---|
| Free | Not included | "Limited" | Basic weekly limits (developer section of Google's plan page) |
| About $20 | Included on Pro | Included on Plus, with the Work and Codex estimates above | AI Pro: "entry" rate limits; Argon not included |
| Top tier | Max: 5x or 20x Pro usage | Pro: no five-hour limit for now; Ultrafast on Pro 500 only | Ultra: 5x or 20x rate limits, with Gemini 4 Argon |
Table 5Claude Code and Codex come with the $20 plans; Antigravity gets Gemini 4 Argon only on Ultra. Source: each company's plan and pricing pages, checked October 20, 2026.
Last verified: October 20, 2026
Claude Code shares one allowance with Claude chat, and Codex shares one with ChatGPT Work. Google's launch post says Argon reaches Antigravity only "for Google AI Ultra subscribers."
What do independent evaluators say?
Evaluators test models through their APIs, usually at high effort settings, not the consumer apps. They measure different things, so we don't merge them into one ranking.
| Evaluator (date) | What it measures | Claude | OpenAI | |
|---|---|---|---|---|
| LMArena Text Arena (snapshot dated October 15, 2026) | Crowd votes on paired chat answers; scores with style control applied | Opus 5.5 (High) 3rd, 1508 ±7 | Best GPT-6 model: GPT-6.1 Sol (Max) 22nd, 1484 ±8 | Gemini 4 Argon (High) 1st, 1524 ±9, marked preliminary |
| Artificial Analysis Intelligence Index v4.3.2 (pages checked October 20, 2026) | Composite of the firm's chosen evaluations | Opus 5.5 ("Max, Default Fallback") 58 | GPT-6 Astra (Max) 53 | Gemini 4 Argon (High) 53 |
| Vals Index v2.2 (updated October 19, 2026) | Agentic tasks in finance, coding, legal and tax | Sonnet 5.5 70.85% (first), Opus 5.5 70.18% | GPT-6 Astra 65.02%; GPT-6.1 Sol 63.94% | Gemini 4 Argon 69.24% |
Table 6The evaluators disagree: Gemini 4 Argon leads LMArena's October 15 snapshot, Claude Sonnet 5.5 narrowly leads Vals AI's October 19 index and Claude Opus 5.5 leads Artificial Analysis's index v4.3.2. Results as published by each evaluator, on the dates shown.
Last verified: October 20, 2026
Limits: Arena voters judge single chat answers, and Argon's lead rests on fewer than 5,000 votes. Artificial Analysis's pages carry no data date. Vals's top three are within two points, and its Claude scores include some tasks a fallback model answered after a provider refusal; counting those as failures, Vals says, would drop Sonnet 5.5 to 69.03% and Opus 5.5 to 64.16% and put Argon first. Vals moved to version 2.2 of its index on October 19, so its scores aren't comparable with earlier versions. For the full model-level comparison, including vendor benchmarks, see Gemini 4 Argon vs Claude Opus 5.5 vs GPT-6.
Which should you pick?
| If you mainly… | Pick | Why |
|---|---|---|
| Want a free assistant without ads | Claude Free or the Gemini app | OpenAI documents ads on ChatGPT Free unless you accept lower limits; the other two mention none |
| Want the cheapest paid plan | Google AI Plus ($4.99) | 128K context and 2x free usage; ChatGPT Go ($8) may show ads |
| Read long PDFs, books or contracts at the $20 level | Claude Pro or Google AI Pro | Both list 1M tokens; ChatGPT Plus lists 256K at most |
| Code with an agent a few hours a week | Claude Pro or ChatGPT Plus | Claude Code and Codex are included; Antigravity with Argon needs Ultra |
| Code with agents all day | Claude Max 5x or ChatGPT Pro 100 | Both $100; Pro has no five-hour limit for now, Max has session and weekly limits |
| Work inside Gmail and Docs | Google AI Pro | Google's plan page lists Gemini in Gmail (from AI Plus) and Docs (AI Pro), plus 5 TB of storage |
| Want Gemini 4 Argon, the top model on LMArena's October 15 text board | Google AI Pro or Ultra | Argon isn't on the free plan or AI Plus |
| Want Claude Opus 5.5, the top model on Artificial Analysis's index v4.3.2 | Claude Pro or above | Opus isn't on Claude Free |
| Want the fastest GPT-6 Astra | ChatGPT Pro 500 ($500) | The only individual plan with Ultrafast |
| Want the least data used for training without a business plan | Claude | Training is opt-in; ChatGPT and Gemini can use chats for training unless you turn the setting off |
| Buy for a small team | Claude Team or ChatGPT Business | Matching $25 monthly / $20 annual seats; compare admin features |
Table 7Our picks by job, based on the prices, limits and published results above; no single plan wins every row.
If you work mostly through an API rather than an app, the trade-offs are different. Our cheap, fast models comparison covers Claude Haiku 5.5, GPT-6 Luna and Gemini Flash at API prices, and our Haiku 5.5 story covers Anthropic's smallest model.
What's confirmed, reported and uncertain
- Confirmed (official pages, October 20, 2026): prices, billing rules, models per plan, context and file limits, usage-limit mechanics, ad eligibility, data settings and coding-tool inclusion.
- Third-party, with limits: the LMArena, Artificial Analysis and Vals AI results, each tied to its date and configuration.
- Uncertain or not published:
- which Claude app plans include Haiku 5.5 (the Help Center's October 7 note doesn't say)
- message counts for Claude plans and for ChatGPT's Chat tab
- Claude's Free-plan context window
- Gemini 4 Argon usage limits per plan in the Gemini app
- local-currency prices, which vary by country
