ChatGPT vs Claude vs Gemini (2026): Plans, Message Limits, and Models Compared Side by Side

The Short Answer

All three run a free tier and a paid tier near $20/month. The differences that actually change your day:

• ChatGPT has the widest ladder of plans (Free, Go $8, Plus $20, Pro $100, Pro Max $200) and the most generous free message count: 10 GPT-5.5 messages every 5 hours.
• Claude bundles Claude Code and Claude Cowork into Pro at $20/month, and lets you turn off model training to keep a 30-day data retention window instead of up to 5 years.
• Gemini is cheapest to upgrade ($4.99 AI Plus), ships a 1M-token context on AI Pro, and includes the most cloud storage (5 TB on AI Pro).

If you write code, Claude leads the published coding benchmarks and bundles a terminal agent, though Fable 5 and Mythos 5 are currently suspended (see note above). If you want the cheapest paid step up, Gemini AI Plus is $4.99. If you want the most free usage, ChatGPT Free gives the clearest message allowance.

Every Plan and Price

ChatGPT Go is $8/month, available in 98 countries (including the EU) since January 2026, and ad-supported. ChatGPT Pro at $100/month gives 5x the usage of Plus and 125 Deep Research queries per month; Pro Max at $200/month gives 20x usage, GPT-5 Pro, and unlimited audio/video mode. Google dropped its top AI Ultra price from $250 to $200/month while keeping the same 20x usage limit. Claude Max plans are monthly only with no annual discount.

Message and Usage Limits

Headline price tells you almost nothing about how much you can actually use each app. The limits do.

ChatGPT publishes the clearest numbers: Free gets 10 GPT-5.5 messages every 5 hours, Plus and Go get 160 every 3 hours (then chats fall back to the mini model until reset), and Plus or Business users can manually select GPT-5.5 Thinking up to 3,000 messages per week. Claude uses a weekly usage limit across all models plus a 5-hour session limit, both viewable in Settings, with Max plans adding a separate weekly cap for Sonnet models. Gemini uses compute-based limits that factor in prompt complexity, features used, and chat length; they refresh every 5 hours until you hit the weekly cap, after which you shift to smaller models.

Which Model Each App Runs

ChatGPT runs GPT-5.5, the latest GPT model in the app, with GPT-5.5 Thinking and GPT-5.5 Pro variants for harder work. Claude Fable 5 launched June 9, 2026 as Anthropic's most capable widely released model, built for demanding reasoning and long-horizon agentic work (currently suspended, see note above); it is the same underlying model as Mythos 5 with added cybersecurity and biology safeguards that route those queries to Opus 4.8. Gemini 3.1 Pro is Google's current flagship with a context window up to 1M tokens, about 1,500 pages of text or 30,000 lines of code; the free tier uses Gemini 3.5 Flash with varying access to 3.1 Pro.

Benchmarks

Each vendor publishes its own numbers on its own scaffold, and the benchmark variants differ, so read these as directional rather than a head-to-head leaderboard.

Anthropic states Claude Fable 5 (currently suspended, see note) is state-of-the-art on nearly all tested capability benchmarks; the SWE-bench and GPQA figures above are the published Opus 4.5 numbers, since Fable 5 just launched. GPT-5.5 leads command-line and agentic coding workflows (82.7% Terminal-Bench 2.0). Gemini 3.1 Pro more than doubles Gemini 3 Pro's ARC-AGI-2 reasoning score, reaching 77.1%.

Does Your Data Train the Model

This is the setting most people never check, and the default differs across all three.

ChatGPT's "Improve the model for everyone" toggle is on by default for consumer accounts; switching it off in Settings means new conversations are not used to train models, and the change applies account-wide and can be reversed anytime. Gemini trains by default via the Keep Activity setting; turning it off stops future chats from being used to improve Google AI (audio, Live videos, and screenshares are already excluded by default). Claude makes model improvement a user choice on Free, Pro, and Max: leave it off and Anthropic does not use your chats for training and applies a 30-day retention window; opt in and data may be retained de-identified for up to 5 years.

Pick by Use Case

Building on These Models Yourself

The consumer apps wrap a model in a chat box. If you are building a product on top of these models or running open-source models like DeepSeek, the chat subscription is the wrong unit; you want an API and per-token pricing. Morph serves open-source and coding-tuned models with full 16-bit activations (no fp8 quantization) so output matches the reference weights, plus codegen-specific speculative decoding for fast apply. See Morph Open Source Models and pricing for the per-token rates.

Frequently Asked Questions

Which is cheapest: ChatGPT, Claude, or Gemini?

All three have a free tier. The cheapest paid plan is Google AI Plus at $4.99/month, then ChatGPT Go at $8/month. The standard paid tiers are within a dollar: ChatGPT Plus $20, Claude Pro $20 ($17 annual), Google AI Pro $19.99.

How many messages do you get on the free plans?

ChatGPT Free allows 10 GPT-5.5 messages every 5 hours, then drops to a smaller model. Gemini Free uses compute-based limits that refresh every 5 hours until a weekly cap. Claude Free has session and weekly usage limits that vary by demand.

Which model does each chatbot run?

ChatGPT runs GPT-5.5. Claude Fable 5 launched June 9, 2026 (currently suspended, see note above). Google AI Pro runs Gemini 3.1 Pro with a 1M-token context; the free tier uses Gemini 3.5 Flash.

Does ChatGPT, Claude, or Gemini train on my conversations?

ChatGPT and Gemini train by default (opt out in Settings > Data Controls and by turning off Keep Activity, respectively). Claude makes it a user choice; left off, it does not train on your chats and keeps a 30-day retention window.

Which is best for coding?

Claude leads published coding benchmarks (Opus 4.5 at 80.9% SWE-bench Verified) and bundles Claude Code. GPT-5.5 scores 58.6% on SWE-Bench Pro and ships Codex. For running open-source coding models on an API, see Morph Open Source Models.

Related comparisons