Brass

Is there a pay-as-you-go version of ChatGPT?

Not for the chat app. ChatGPT is sold as monthly plans, Free, Go at $8, Plus at $20 and Pro at $100, $200 or $500, with no per-message option. Pay-per-use exists in two places: the OpenAI API, which bills by the token and requires an API key, and ChatGPT credits, which OpenAI calls a pay-as-you-go add-on but which only cover Codex, ChatGPT Work and the Office add-ins after a plan’s included usage runs out.

Plan prices come from OpenAI’s pricing documentation, checked October 2, 2026. The API prices below come from the API pricing page on the same day.

The three ways to pay OpenAI

Subscriptions. Plus is $20 a month, billed monthly only; OpenAI does not offer annual billing for Go, Plus or Pro. The ChatGPT pricing page lists everyday text chats as unlimited on every plan, including Free, subject to reasonable use, with the paid tiers adding more image generation, uploads, deep research, and Work and Codex allowance. The Pro tiers are $100, $200 and $500 a month, and only Pro 500 includes GPT-6 Astra Ultrafast.

The API. Prepaid API billing means you buy credits first, with a $5 minimum, and usage is deducted from them. Credits expire one year after purchase and are non-refundable. You can turn on auto-recharge and cap it per month. Using the API means an API key plus either your own code or a third-party app that accepts the key; OpenAI’s Codex documentation lists an API key as a way to run Codex in the CLI or IDE extension at API prices with no plan at all.

ChatGPT credits. OpenAI’s credits article describes credits as a pay-as-you-go add-on for supported features once you hit plan limits. The supported features are Codex, ChatGPT Work and ChatGPT for Excel, and the Pro tiers article adds Word and PowerPoint. They do not buy more ordinary chat messages. Credits are valid for 12 months from purchase, non-refundable except where the law requires, non-transferable, and separate from API credits and from the gift-card wallet. They are metered in credits per million tokens: 50 for input and 250 for output on GPT-6.1 Sol, 250 and 1,250 on GPT-6 Astra, 2.5 and 12.5 on GPT-6 Luna. OpenAI does not publish a fixed dollar price per credit; it says purchase prices depend on your plan.

What one message costs on the API today

OpenAI’s token guide gives a rule of thumb: 100 tokens is about 75 English words. Take a 150-word question and a 450-word answer, about 200 input tokens and 600 output tokens, at list prices on October 2, 2026.

Model Input per 1M tokens Output per 1M tokens This exchange
GPT-6 Astra $10.00 $50.00 about 3.2 cents
GPT-6.1 Sol $2.00 $10.00 about 0.6 cents
GPT-6 Luna $0.10 $0.50 about 0.03 cents

Three things push the real number up. Reasoning tokens, the model’s internal working before the visible answer, are billed as output tokens, so a short answer can cost more than its length suggests. Chat history is re-sent as input on every turn; OpenAI’s pricing documentation lists the prompt, files, chat history, tool results and the response as things that use tokens. By the time a thread holds 20,000 tokens of history, each new turn on GPT-6.1 Sol costs 4 cents in input before the answer is written. Cached input is discounted (GPT-6.1 Sol charges $0.10 per million cached input tokens), which softens this for repeated context. Batch processing at half price exists, but it is for jobs you submit and collect later, not live chat.

Taken one exchange at a time, $20 of API credit buys roughly 3,000 short exchanges on GPT-6.1 Sol or 600 on GPT-6 Astra. In a long conversation it buys far fewer.

Pay-as-you-go versus Plus

For someone who uses ChatGPT every day, Plus at $20 is cheaper than paying per message, because text chat on Plus is unlimited within reasonable use and the plan adds a Work and Codex allowance that OpenAI estimates at 15 to 160 local GPT-6.1 Sol messages per five-hour window. For someone who asks a few questions a week, $5 of API credit can last months, with two catches: the credit expires after a year, and the API needs a client to chat through.

The phrase pay-as-you-go also hides a prepaid structure. OpenAI’s API billing is prepaid: you load money, and when the balance hits zero, requests fail with a credit_balance_exhausted error until you add more, per the spend limits guide. Accounts on monthly invoicing are billed for usage beyond their credits instead. ChatGPT credits are prepaid too. The one consumer option billed after use is the monthly subscription itself.

Brass takes the prepaid route without the expiry: packs of $10 to $250 that never expire, spent in a web chat across Claude, GPT, Gemini, Grok and DeepSeek at each model’s list price plus 25 percent, with a receipt on every message and no API key.

If you already have a plan and hit the wall

Plus and Pro users who reach a Work or Codex limit can buy credits in Settings, then Usage, or an instant reset that restores the five-hour and weekly allowance at once, per OpenAI’s reset article. Neither helps with the ordinary chat models, where the options are to wait for the reset time shown in the app, switch models or upgrade.

The article ChatGPT says you reached your limit. What can you do? walks through each limit and its reset clock.

Sources

  1. OpenAI: ChatGPT and Codex pricing
  2. OpenAI API pricing
  3. OpenAI: Setting up and managing prepaid API billing
  4. OpenAI: Using credits for flexible usage in ChatGPT (personal plans)
  5. OpenAI: What is ChatGPT Plus?
  6. OpenAI: About ChatGPT Pro tiers
  7. ChatGPT pricing
  8. OpenAI: Understanding and counting tokens
  9. OpenAI: Spend limits
  10. OpenAI: Paid weekly Work and Codex rate limit resets

Related guides