DEEIX Chat billing is not a single price table. It supports self mode, period billing, usage billing, free models, token pricing, per-call pricing, per-second pricing, tiered pricing, cache billing, service multipliers, native tool billing, and redemption codes.
Billing Structure#
Think about billing in three layers:
| Layer | Managed Content | Impact |
|---|---|---|
| Billing mode | Self, period, usage | Decides whether users see subscription, balance, or usage records only. |
| Model pricing | Token, per-call, per-second, tiered, free | Decides how each model call is priced. |
| Extra billing | Cache, service multipliers, native tools | Decides whether additional items are priced separately or by multiplier. |
Choose billing mode first, then configure model prices and extra billing. Do not change prices without confirming whether users are on period or usage billing.
Three Billing Modes#
Self mode records usage only. Users do not see purchase, top-up, or plan deduction. It fits internal deployments, shared team cost, or trial phases.
Period billing grants period credit through subscription plans. Users buy a plan and calls consume the period credit. It fits monthly credit, yearly credit, and fixed entitlement models.
Usage billing deducts from account balance. Users top up or redeem balance, and calls consume balance according to the real bill. It fits variable usage and budget control by actual consumption.
Prepaid Amount#
Usage billing can require a prepaid amount before a call starts. This is not the final cost; it is a balance threshold.
The threshold reduces failure from insufficient balance during long tasks. Expensive models, long context, and multi-tool calls benefit from a reasonable threshold, but a threshold that is too high makes usage harder to start.
Model Pricing Modes#
| Pricing Mode | Basis | Best For |
|---|---|---|
| Token | Input, output, cache read/write | Mainstream text and multimodal models. |
| Per call | Fixed cost per call | Fixed-cost model operations or tool-like models. |
| Per second | Duration | Video, audio, or time-sensitive generation. |
| Tiered | Different ranges by original input size | Long context, large input, or provider tiered prices. |
Models can also be marked free. Free models are not separately billed and can be described as not consuming period credit.
Token Pricing#
Token pricing can separately set input, output, cache read, and cache write prices. This matches real provider pricing better than one flat token price.
Input and output are the common items. Cache read and write are important for models with cache capability, especially long-context, repeated-context, and Claude-style caching.
Cache Billing#
Cache billing handles provider differences for cache read, cache write, and cache write duration such as 5 minutes or 1 hour.
For example, Claude cache write can use configured price multiplied by different factors. Fast Mode may apply another multiplier to input, output, and cache costs. This keeps user display simple while preserving real cost structure in admin and logs.
Cache billing matters for long conversations and large documents. Without cache billing, similar token counts can be hard to explain when costs differ.
Service Multipliers#
Some providers apply multipliers for service tier, fast mode, or priority mode. DEEIX Chat can reflect these multipliers in billing explanation and call details.
Administrators should keep pricing notes clear: which models or modes create multipliers, and which billing items are affected. If users show cost metadata, call details should reflect realistic cost.
Tiered Pricing#
Tiered pricing charges different ranges based on original input size. It fits providers that price long context or large input by level.
Define each tier limit and price clearly. The final tier can be unlimited. Too many tiers are hard to maintain; too few may not reflect cost differences accurately.
Native Tool Billing#
Provider-native tools can be billed per call. Tool price and model price are separate: the tool call is priced as a tool, while model input and output around the tool follow model pricing.
A tool price of 0 means no separate tool charge, only usage record. For provider tools with extra cost, such as search, execution, browsing, or external capability tools, configure separate tool prices.
Plans and Redemption Codes#
Subscription plans include price, interval, credit, and discount. After purchase, users receive entitlement segments; renew, upgrade, and switch actions may take effect immediately or be queued depending on existing entitlements.
Redemption codes can grant balance or plan entitlements. They can be limited by billing mode, expiration, total usage, per-user usage, and account scope. They are useful for campaigns, internal grants, compensation, and manual activation.
Billing Snapshots and Usage Logs#
Usage logs record the billing result and key context for each call. Price changes usually affect future calls only; historical calls keep the billing snapshot from their time.
When investigating cost, do not look only at the current price table. Open usage log details and check model, tokens, cache, multipliers, tool calls, and billing mode.
Configuration Advice#
Use self mode while validating models and file features, then switch to period or usage billing. Before launch, configure prices for core models, free models, cache, and native tools. After every price change, make a test call and check user-side cost, subscription page, usage logs, and admin billing.