AI Models
Configure which AI models your organization uses, understand the base, low and image models, compare provider pricing, and learn how AI credit usage is tracked and billed across all Yaplet features.
The Model Slots
Base Model
Handles complex, resource-intensive tasks that require deep reasoning and longer outputs. Used for chatbot answer generation, AI suggestions, and content creation.
Low Model
Handles fast, lightweight tasks that need quick turnaround. Used for grammar correction, message rephrasing, conversation summaries, and supporting operations that make the base model more efficient.
Image Model
Generates images — for example the visuals of AI social posts. Billed at a flat price per generated image rather than per token.
Which features use which model?
| Feature | Model | Examples |
|---|---|---|
| Vex chatbot streaming answers | Base | Generating real-time responses to visitor questions |
| AI content generation | Base | Writing knowledge base articles, generating suggestions |
| Inbox AI Tools | Low | Grammar check, rephrase, translate |
| Conversation summaries | Low | Summarizing chat history for context |
| Searching the brand's knowledge for an answer | Low | Searching everything the brand owns in one pass — knowledge bases, documentation, products, and the files, text, Q&A and website pages added to it |
| Image generation | Image | Cover visuals for AI social posts |
Configuring Your Models
Navigate to Settings → Organization settings → AI Models to view and change your organization's model configuration.
Current Models
The settings page displays your active models, including their provider name and per-million-token pricing for all cost types:
- Input — cost per million input (prompt) tokens
- Output — cost per million output (response) tokens
- Cached Input — discounted rate for cached prompt tokens
- Reasoning — cost for advanced reasoning tokens (used by models like OpenAI o-series or Claude with extended thinking)
- Per image — flat cost per generated image, shown for the image model instead of token prices
Each model shows a badge indicating whether it's the Yaplet Default or a Custom selection.
Changing Models
Select new models
Use the dropdown menus to pick a new base model, low model, image model, or any combination. Models are grouped by provider (OpenAI, Anthropic, Google, xAI) with pricing shown for each option.
Review the cost comparison
When you select a different model, a comparison table appears showing current vs. selected pricing for each cost type. Increases are highlighted in red, decreases in green — so you can see the cost impact at a glance.
Confirm and apply
Click Update Models and confirm the change. The new models take effect immediately for all AI tasks across your organization.
Supported Providers
Yaplet integrates with leading AI providers through a unified interface. You don't need API keys or separate accounts — all provider access is managed through your Yaplet subscription.
| Provider | Strengths |
|---|---|
| OpenAI | Broad model range, strong general-purpose performance |
| Anthropic | Advanced reasoning, large context windows |
| Competitive pricing, multimodal capabilities | |
| xAI | Fast inference, latest research models |
Credit Usage & Billing
Every AI operation consumes tokens — small units of text that models process. Yaplet tracks usage across all features and bills using a unified currency called Yaplet Credits.
How credit billing works
Tokens are consumed
Each AI request uses input tokens (your prompt) and output tokens (the model's response). Some models also use cached input tokens and reasoning tokens.
Costs are normalized
Because different token types have different prices, all usage is converted to input-token equivalents using the model's pricing ratios. This gives you a single, comparable number.
Billed as Yaplet Credits
The normalized usage is converted to Yaplet Credits at a rate of $0.001 per credit ($10 per 10,000 credits). This unified pricing makes it easy to understand costs regardless of which model or provider you use.
Token cost types explained
| Token Type | What It Is | When It Applies |
|---|---|---|
| Input | Tokens in your prompt and context | Every request |
| Output | Tokens the model generates | Every request |
| Cached Input | Repeated prompt tokens served from cache (lower cost) | When conversation context is reused |
| Reasoning | Internal thinking tokens used for complex problem-solving | Models with reasoning capabilities (e.g., OpenAI o-series) |
Image generation is the exception: it isn't measured in tokens at all. The image model carries a flat cost per image, shown on the settings page alongside the other models.
Trial allowance vs. the free plan
These are two different things, and it's easy to mix them up. A 14-day trial starts with a paid subscription and includes 500 AI credits and 1,000 emails; using up either allowance ends the trial early and your paid subscription starts billing straight away — the owners get an email saying so. The free plan is separate: it includes 100 AI answers per organization per month, after which answers are paid from your prepaid credit balance at the standard rate. See Plans & Pricing for details.
Where AI Is Used in Yaplet
AI models power features across the entire platform. Here's a quick overview of where your model choices have an impact:
Vex AI Chatbot
Your base model generates real-time streamed answers to visitor questions, using the brand's knowledge, products and API tools.
Inbox AI Tools
Your low model powers grammar checking, rephrasing, and other message enhancement tools available to agents in the inbox.
Tips for Choosing Models
If budget is a priority, choose models with lower per-token pricing — especially for the low model, which handles high-volume simple tasks. A cheaper low model can significantly reduce your overall token spend.
For the best chatbot responses, select a premium base model with strong reasoning capabilities. The base model directly affects the quality of answers your visitors receive.
If you're unsure, keep every model set to Use Yaplet's Default. Yaplet regularly evaluates models across providers and updates the defaults to offer the best balance of quality and cost.