[IMAGE: Model dropdown expanded showing options]
Location
Prompt Section (top bar) → First dropdownAvailable Model Types
| Type | Description |
|---|---|
| Emotive Models | Speech-to-speech with emotional tone and expression (Beta) |
| Traditional Models | Standard LLM models for text processing |
Available Models
| Model | Provider | Best For |
|---|---|---|
| GPT-4o | OpenAI | Complex conversations, nuanced understanding |
| Electron | Atoms | Speed and efficiency |
How to Select
- Click the model dropdown in the Prompt Section
- Browse available options
- Click to select your choice
Choosing the Right Model
GPT-4o (Recommended for most)
- Best understanding and reasoning
- Handles complex queries well
- Slightly higher latency than smaller models
- Recommended for support, sales, complex workflows
Emotive Models (Beta)
- Natural emotional expression
- More human-like responses
- Good for conversations requiring empathy
- Beta — may have quirks
Electron Models
- Optimized for speed
- Lower latency
- Good for simple, high-volume use cases
- Cost-effective
Considerations
Latency
Larger models take longer to respond. For voice AI, every millisecond matters:- Simple queries → smaller models may be sufficient
- Complex reasoning → larger models worth the tradeoff
Cost
Different models have different per-minute costs:- Check your plan details for pricing
- Higher-capability models typically cost more
Accuracy
More capable models generally:- Follow instructions better
- Handle edge cases more gracefully
- Provide more nuanced responses
Best Practices
- Start with GPT-4o for most use cases
- Test thoroughly when changing models
- Compare latency in real calls
- Review conversation logs to verify quality