Managing AI Models
As an agency, you control which of the 120+ AI models your clients can access. This guide explains how to enable models and set your pricing.
Accessing Model Management
Navigate to Dashboard → Models & Providers (agency admins only)
You'll see a table of all available models from AI Gateway with:
- Provider (OpenAI, Anthropic, Google, etc.)
- Model name and capabilities
- Input/output pricing
- Your pricing markup
- Enable/disable toggle
Understanding the Pricing Flow
Models have a three-tier pricing structure:
AI Gateway Base Cost
↓ + Platform Markup (20% default)
Platform Cost (what you pay)
↓ + Your Agency Markup (you set this)
Client Cost (what clients pay)
Example:
- Gateway input: $0.00250 per 1K tokens
- Platform markup: 20%
- Your cost: $0.00300 per 1K tokens
- Your markup: 50%
- Client pays: $0.00450 per 1K tokens
- Your profit: $0.00150 per 1K tokens
Enabling Models for Clients
Step 1: Browse Available Models
The Models page shows all 120+ models grouped by provider:
- OpenAI - GPT-4o, GPT-4o-mini, GPT-3.5-turbo
- Anthropic - Claude 3.5 Sonnet, Claude 3 Opus, Claude 3 Haiku
- Google - Gemini Pro, Gemini Flash
- Alibaba - Qwen models
- Meta - Llama models
- And many more
Step 2: Enable a Model
- Find the model in the table
- Toggle the Enable switch to ON
- Model is now available to all your clients
Step 3: Set Your Markup (Optional)
By default, you charge clients the same price you pay (0% markup). To add your margin:
- Click the markup percentage field
- Enter your markup (e.g., 50 for 50%)
- Save changes
Tip: Most agencies use 30-50% markup for sustainable margins
Step 4: Mark as Recommended (Optional)
Highlight models you suggest:
- Toggle Recommended switch
- These appear first in dropdown lists for clients
Choosing Which Models to Enable
Start with These
For most agencies, enable these 3 models initially:
1. gpt-4o-mini (OpenAI)
- Fast and cost-effective
- Great for customer support, simple tasks
- Input: ~$0.003/1K tokens
2. gpt-4o (OpenAI)
- More powerful for complex reasoning
- Good for sales, detailed analysis
- Input: ~$0.015/1K tokens
3. claude-3-5-sonnet (Anthropic)
- Excellent for nuanced, detailed responses
- Great for content writing, research
- Input: ~$0.018/1K tokens
Add More As Needed
Enable additional models when clients need:
- Cheaper options - GPT-3.5-turbo, Claude 3 Haiku
- More power - Claude 3 Opus, GPT-4-turbo
- Specific features - Gemini for Google integration
Model Selection by Use Case
Customer Support
- Best: gpt-4o-mini
- Why: Fast, accurate, cost-effective
- Cost: Low (~$0.003/1K input tokens)
Sales & Persuasion
- Best: gpt-4o or claude-3-5-sonnet
- Why: Better at nuance and persuasion
- Cost: Medium (~$0.015-0.018/1K input tokens)
Content Writing
- Best: claude-3-5-sonnet
- Why: Creative, detailed, nuanced
- Cost: Medium (~$0.018/1K input tokens)
Complex Analysis
- Best: gpt-4o or claude-3-opus
- Why: Advanced reasoning capabilities
- Cost: Medium to High
Budget-Conscious
- Best: gpt-3.5-turbo or claude-3-haiku
- Why: Cheapest options, still capable
- Cost: Very low (~$0.001/1K input tokens)
Setting Your Markup Strategy
Markup Recommendations
Conservative (20-30%)
- Good for competitive markets
- Lower client costs = easier sales
- Smaller margins per usage
Moderate (40-50%)
- Balanced approach (recommended)
- Sustainable margins
- Still competitive pricing
Premium (60-100%)
- For high-touch service
- When you provide significant value-add
- White-label with premium positioning
When to Adjust Markup
Increase markup when:
- Providing extensive setup/support
- Client has high usage volume (better margins)
- You're bundling other services
Decrease markup when:
- Testing new clients (introductory pricing)
- Competing for large contracts
- Client is price-sensitive
Disabling Models
If a model isn't working well or costs too much:
- Toggle the model to OFF
- Existing agents using this model will fallback to tenant default
- Clients won't see it in model dropdown anymore
Tip: Don't disable models that agents are actively using without warning clients
Client View vs Agency View
Clients see:
- Only enabled models
- Final price (after your markup)
- Model capabilities and descriptions
Clients don't see:
- Gateway base pricing
- Platform markup percentage
- Your markup percentage
- Disabled models
Best Practices
✅ Do:
- Start with 3-5 essential models
- Set consistent markup across similar models
- Mark recommended models for easier selection
- Monitor which models clients use most
❌ Don't:
- Enable all 120+ models at once (overwhelming)
- Set different markups randomly (confusing)
- Disable models without warning
- Forget to check pricing changes from gateway
Cost Control Tips
Monitor usage:
- Check which models are most used
- Identify expensive models with low value
- Disable unused models
Guide clients:
- Recommend cost-effective models
- Show them usage analytics
- Suggest cheaper alternatives when appropriate
Set expectations:
- Explain model pricing upfront
- Show cost estimates for typical usage
- Help clients budget appropriately
Troubleshooting
Model not appearing in agent creation?
- Ensure model is enabled (toggle ON)
- Check client is in correct workspace
- Refresh the page
Pricing seems wrong?
- Verify platform markup percentage in agency settings
- Check your agency markup in models table
- Gateway pricing may have changed
Too many models overwhelming clients?
- Enable only 3-5 essential models
- Use "Recommended" flag to guide choices
- Add descriptions in client onboarding
Next Steps
Now that your models are configured:
- Credits & Billing - Manage your balance
- Team & Permissions - Add team members
- Analytics & Monitoring - Track usage