Model Selection
Last updated: Jan 2026
Overview
Choosing the right AI model is crucial for workflow success. Different models excel at different tasks, and the best choice depends on your specific requirements for quality, speed, and cost.
Capability
What it can do
Speed
Response time
Cost
Credits per token
Accuracy
Output quality
Available Models
ORCFLO supports models from leading AI providers. Each offers different strengths and capabilities.
Anthropic Claude
200K context window
Excellent for nuanced reasoning, long-form content, and tasks requiring careful analysis. Strong safety features and high-quality outputs.
OpenAI GPT
Up to 1M context window
Versatile models with strong general capabilities. Includes reasoning models (o3, o4-mini) for complex problem-solving. Good for function calling and multimodal tasks.
Google Gemini
Up to 1M context window
Fast and efficient models with massive context windows. Great for high-volume processing and long document analysis. Cost-effective for large-scale operations.
Selection Criteria
Consider these factors when choosing a model for your workflow.
| Factor | Choose Larger Model | Choose Smaller Model |
|---|---|---|
| Task Complexity | Complex reasoning, nuanced analysis | Simple classification, extraction |
| Quality Requirements | Customer-facing, high stakes | Internal tools, acceptable errors |
| Volume | Low volume, high value | High volume, cost-sensitive |
| Latency | Batch processing, async | Real-time, user-facing |
| Context Length | Long documents, conversations | Short inputs, simple prompts |
Task Recommendations
Here are recommended models for common workflow tasks.
| Task | Recommended Models |
|---|---|
| Content Generation | Claude Sonnet 4.5 (quality), GPT-5.2 (versatility), Gemini 2.5 Pro (speed) |
| Data Extraction | Claude Haiku 4.5 (cost), GPT-4o Mini (fast), Gemini 2.5 Flash (volume) |
| Code Tasks | Claude Sonnet 4.5 (accuracy), GPT-5.2 (broad support), Gemini 2.5 Pro (speed) |
| Classification | Claude Haiku 4.5 (efficient), GPT-4o Mini (reliable), Gemini 2.0 Flash (fast) |
Start Small, Scale Up
Begin with a smaller, faster model and only upgrade if quality isn't meeting requirements. Many tasks don't need the most powerful model.
Cost Optimization
Model costs can add up quickly. Here are strategies to optimize spending.
- Right-size your model: Use smaller models for simpler tasks. Not every step needs Claude Opus 4.5.
- Minimize tokens: Write concise prompts where possible.
- Batch processing: Process multiple items in single requests when supported.
Configuring Models
ORCFLO makes it easy to configure models for each LLM step. You can search, filter by pricing tier, and fine-tune parameters.
Open the model selector
Click the current model name displayed on your LLM step to open the model selector.
Search or filter by tier
Choose from Lite, Value, or Premium tiers to narrow your options by cost and capability.
Select the model
Click on your desired model to apply it to the step. The selector will close automatically.
Open the LLM node configuration panel
Click the 'Settings' button on the LLM step to access advanced configuration options.
Adjust temperature
Lower values (0.0–0.3) produce focused, deterministic outputs. Higher values (0.7–1.0) increase creativity and variation.
Test with sample data
Run the workflow with representative inputs to verify output quality before committing to the new model.
Monitor changes
Track cost and step duration differences in the execution pane to ensure the model meets your requirements.
Prompt Adjustments
Some prompts work better with specific models. When switching, you may need to adjust your prompts for optimal results.
Best Practices
- Match model capability to task complexity
- Use smaller models for high-volume, simple tasks
- Reserve powerful models for complex reasoning
- Test multiple models before committing
- Monitor costs and quality metrics
- Consider latency requirements for real-time use
- Keep prompts model-agnostic when possible
Key Takeaways
Different models excel at different tasks - match model to requirement. Consider capability, speed, cost, and accuracy trade-offs.
Start with smaller models and upgrade only if needed. Use task-specific recommendations as starting points.
Monitor and optimize costs with right-sizing. Test model switches with sample data before production.