Most AI products charge a flat monthly fee. To make the math work, providers have to keep average compute per user below the subscription price. That means silent trade-offs: routing to smaller models (the AI systems that generate responses), truncating context (how much of the conversation and attached documents the model can see), or throttling during peak demand.
Apers charges by actual compute instead. We call the units Smart Request Credits.
How credits work
AI models process text in tokens, small units, typically a few characters each. Each request consumes tokens across input (your message and documents), output (the response), and cache (reused context from prior turns). Each type has a different per-token rate.
Your credit charge comes from the actual token counts and rates for that request. Fractional credits are tracked precisely; your dashboard rounds for readability.
What drives cost
Query complexity. A short question costs well under 1 credit. Comparing five documents costs more. Building a multi-tab spreadsheet with linked analysis costs more still. Cost scales linearly with tokens.
Model choice. The default model handles most tasks well. Max Mode uses a more powerful model for deeper reasoning and costs more per token. Eco Mode uses a lighter model at lower cost and is available when you've run out of credits.
Context size. Very long documents cost more to process than short ones, since the model has to read more input.
Caching
Apers automatically reuses context from prior conversation turns. Cached tokens cost roughly 90% less than fresh input. This adds up in multi-turn conversations and document-heavy workflows.
Credits and pricing
Credits are included with your subscription plan and refresh monthly. If you need more, you can buy additional credit packages at any time; larger packages come at a lower per-credit rate.
New accounts start with 25 free credits to try things out.
