Side by side

ChatGPT vs Claude: price, limits, context and API rates

ChatGPT Plus and Claude Pro both cost $20.00 a month. Their published limits, context windows and separate API rates differ, and every figure below comes from the same checked ledger as their product pages.

Ledger rows last checked 2 Nov 2026.

ChatGPT Plus and Claude Pro, using only figures their vendors publish
Measure	ChatGPT Plus	Claude Pro
Monthly list price	$20.00 a month. Read from learn.chatgpt.com/docs/pricingon 2 Nov 2026.	$20.00 a month. Read from claude.com/pricingon 5 Sep 2026.
Annual price per month	Not published for this individual plan.	$17.00 when paid annually.
Plan context window	Not published. The checked plan source publishes no context-window figure, so this is not treated as zero or estimated from an API model.	1M tokens for the paid plan models named in the checked source.
Published usage limits	GPT-6 Astra local messages: 5-45 per five-hour period. GPT-5.6 Sol local messages: 10-100 per five-hour period. GPT-5.6 Terra local messages: 25-200 per five-hour period. GPT-5.6 Luna local messages: 250-2,000 per five-hour period. GPT-5.5 local messages: 15-80 per five-hour period. GPT-5.4 local messages: 20-100 per five-hour period. GPT-5.4 mini local messages: 60-350 per five-hour period. Weekly limits: may also apply per week.	Usage vs Free: At least five times the usage per session compared to our free service. Session reset: Session-based usage limit will reset every five hours. Weekly limit: Pro plans also have a weekly usage limit that applies across all models per week.
Cheapest API input rate in this ledger	$0.20 per million input tokens for GPT-5.6 Luna; cached input $0.02, output $1.20. Read from developers.openai.com/api/docs/pricingon 2 Nov 2026.	$1.00 per million input tokens for Claude Haiku 4.5; cached input $0.10, output $5.00. Read from platform.claude.com/docs (pricing)on 5 Sep 2026.

Download the ledgers behind this comparisonplans JSON plans CSV pricing JSON pricing CSV

What the comparison says

The monthly sticker price is a tie. It stops being a tie when annual billing, plan context and the wording of the limits matter: Anthropic publishes an annual equivalent and a context window for Claude Pro, while the checked ChatGPT Plus source publishes model-by-model local-message ranges. An unpublished figure is not a smaller figure, and it is never filled from a different product.

The API row is a separate purchase from either subscription. The table compares the cheapest positive dollar input rate each vendor has in this ledger, names the model beside it and keeps the output and cached-input rates attached. It does not imply GPT-5.6 Luna and Claude Haiku 4.5 are equal in capability.