Model Selection
GPT-4.1: Latest flagship, enhanced reasoning. Best for complex analysis and creative tasks. GPT-4.1-mini: Faster, cost-effective. Best for balanced performance and standard processing. GPT-4.1-nano: Ultra-fast, basic operations only. Best for high-throughput simple tasks. GPT-5 / GPT-5-mini / GPT-5-nano: Next-generation models with advanced capabilities.Selection Criteria
Task Complexity: Use nano/mini for simple extraction, full models for nuanced reasoning. Volume and Cost: High volume = mini/nano to control costs. Critical tasks = invest in full models. Response Time: nano (~500ms), mini (~1-2s), full (~2-4s), GPT-5 (~3-6s).API Key Management
Store API keys in workspace variables, never directly in agent configuration. Use separate keys for development and production. Monitor usage and set limits on provider dashboards.Cost Management
Token-Based Pricing: Input tokens (prompt/context) + Output tokens (response) + Tool tokens. Optimization:- Write efficient prompts without unnecessary context
- Use appropriate models (don’t over-engineer simple tasks)
- Include only necessary tools
- Process records in batches
Next Steps
Chat with Agents
Test model configuration interactively
Managing Tools
Configure tools for agents
Creating Agents
Create agents with model configs
Testing Agents
Comprehensive agent testing