By Quad Chat · Published · Last updated
How AI Credits Work in a Multi-Model App
A credit is a single monthly budget that every model draws from. Instead of paying one vendor per subscription, you get a pool of credits and spend it on whichever model you pick: a small, cheap model uses very little, a frontier model uses more. This guide explains how that works in Quad Chat, using the numbers on its pricing page.
Direct answer: In Quad Chat, credits are shared across every model, they reset each billing period, and web searches are counted separately. The same message costs fewer credits on a lightweight model than on a flagship one, so a plan's credits do not map to a fixed number of messages. Free includes 100 credits a month; paid plans run from 600 on Go to 20,000 on Max.
What a credit is
A credit is Quad's internal unit of model usage. It exists because model providers price in tokens, and each provider uses a different rate card. Credits turn that into one number you can budget against, whichever model you choose.
Three rules cover most of the behavior:
- One pool for all models. GPT, Claude, Gemini, Grok, DeepSeek and the rest draw from the same balance. You do not buy separate allowances per provider.
- Cheaper models use fewer credits. The model picker shows an estimate before you send, and Studio shows an estimated credit cost before you generate an image or a video.
- Searches are separate. Web searches have their own monthly allowance and do not come out of your credits.
What each plan includes
These figures come from Quad's plan configuration and match the public pricing page.
| Plan | Monthly price | Yearly, per month | Credits | Web searches | Chat models available |
|---|---|---|---|---|---|
| Free | $0 | n/a | 100 | 100 | 32 of 49 |
| Go | $8 | $7 | 600 | 400 | 32 of 49 |
| Pro | $15 | $13 | 1,200 | 1,000 | 44 of 49 |
| Ultra | $40 | $34 | 2,500 | 4,000 | 48 of 49 |
| Max | $200 | $170 | 20,000 | 20,000 | 49 of 49 |
Yearly billing is 15% less than twelve monthly payments. Plans differ in two ways: how much usage they include, and how much of the model catalog they unlock. Video generation starts on Ultra.
Measured as credits per dollar of monthly price, Go gives 75, Pro 80, Ultra 62.5 and Max 100. That is only one lens: higher tiers also unlock more of the catalog, video generation and more file storage, so compare plans on what you will actually use.
Why credits do not equal messages
Because a credit is a budget rather than a message count, the number of messages you get depends on which models you use. The pricing page gives a sense of scale with one reference model, Claude Sonnet 5.5, assuming a typical message of about 1,500 input tokens and 700 output tokens:
| Plan | About this many Claude Sonnet 5.5 messages a month |
|---|---|
| Free | 55 |
| Go | 500 |
| Pro | 800 |
| Ultra | 2,500 |
| Max | 12,000 |
Treat these as estimates for one model and one message size. Long documents, long answers and reasoning-heavy models use more. Smaller, faster models use far less, so the same plan stretches much further if you reserve flagship models for the work that needs them. The per-token pricing guide shows how much the gap between models can be.
Reading the cost before you send
Quad shows a credit estimate before you commit. In chat, the model picker shows the estimate for the model you selected. In Studio, the generation control shows "about N credits" for the image or video you are about to make. If you compare several models at once, each answer is its own generation, so comparing three models uses more of your allowance than a single answer. The side-by-side guide covers when that is worth it.
What happens when you run out
Running out of credits does not lock you out of your account. You can wait for the next period, upgrade, or buy a one-time credit pack. Unused credits do not roll over between billing periods. Bonus credits earned from referrals are the exception: they never expire.
How team plans work
Team plans use the same idea per seat. Team is $20 per seat a month with 1,500 credits and 2,000 web searches for each member. Business is $40 per seat with 4,000 credits and 5,000 web searches. Each seat has its own private allocation: credits are not pooled across the team, and admins see usage totals but never member conversations or files.
How to make credits last
- Match the model to the task. Use a lightweight model for drafts, rewrites and summaries, and a flagship model for the answer that has to be right.
- Check the estimate on big jobs. Long files and long outputs are where credits go.
- Compare deliberately. Comparing three models is a decision tool, not a default.
- Watch searches separately. If you run a lot of research, the search allowance may run out before credits do.
When separate subscriptions still make sense
A shared credit pool suits people who spread their work across several models. If nearly all of your usage sits inside one provider and you need its native voice, memory or maximum sustained usage, a direct subscription can be the better fit. The one-subscription guide walks through that decision, and ChatGPT Plus vs Quad Pro compares a single provider's plan with Quad's Pro plan.
Summary
Credits are one shared monthly budget: cheaper models spend less, searches are counted separately, and the estimate appears before you send. Pick a plan by how much usage you need and how much of the catalog you want to unlock, not by an imagined message count.
Plan figures were checked against Quad's pricing configuration on October 5, 2026. Prices and allowances can change, so confirm on the pricing page before you subscribe.
See also: AI model pricing per million tokens, explained | How to compare AI models side by side | How to save money on AI subscriptions