Model Usage
Providers & Models
The AI providers PrettiFlow uses and what each is good at.
PrettiFlow can use five AI providers. Each has different strengths, speed, and cost. PrettiFlow chooses among them per task.
Supported providers
| Provider | Best at | Relative cost |
|---|---|---|
| Qwen (default) | General work and code, very economical | Lowest |
| Groq | Fast open-source models | Lowest |
| OpenAI | Reliable structured output and code | Medium |
| Anthropic (Claude) | High-quality reasoning and code | Highest |
| Google Gemini | Google's multimodal models | Highest |
"Relative cost" reflects how much each provider draws from your credits. See
Credit cost per model for the exact weights.
How they're used
- Qwen is the default workhorse for most tasks, including code execution.
- Groq serves fast open models, often for sub-agent work.
- OpenAI is favored where reliable, structured responses matter (and powers code generation when configured).
- Anthropic and Gemini are available for high-quality output where it's worth the extra cost.
The mix of providers available can depend on your deployment's configuration.
PrettiFlow always falls back to an available provider so your build keeps going.