
Evaluate open-source models to reduce API costs in your agent workflows
You'll have a cost-comparison matrix showing token spend and performance trade-offs between your current paid API and 2-3 open alternatives you've tested on your actual workload.
What you'll make
Agent Cost Optimization Matrix
- Current Baseline: API costs and performance
- Open Model Test Results: 3 alternatives with real metrics
- Switching Recommendations: task-specific routing decisions
The 3 parts
Each part ends with a check you can see.
Map Your Current Agent Costs
Track actual API spend and token usage across your agent's different task types
Done when
You have 7+ days of real token usage data broken down by agent task type with actual costs
Test Open Models on Real Tasks
Deploy 2-3 open-source models and run your actual agent tasks through them to measure performance
Done when
You have test results from 3+ open models on your actual agent tasks with quality scores and token counts
Design Your Model Routing Rules
Create specific rules for which tasks go to which models and implement a switching system
Done when
You have documented routing rules for 3+ task types and a working implementation that switches models based on task characteristics
Capstone: you put together your Agent Cost Optimization Matrix.
Where this helps
An elective: a single lesson that sits outside rtwo's job certificates and doesn't count toward one.
About this lesson
What you'll be able to do
- Deploy and test 2-3 open-source models on your actual agent workloads
- Build a cost-performance matrix comparing open vs paid API spend on real tasks
- Make and defend model switching decisions for specific agent workflows
- Implement a multi-provider model switching pattern in your existing setup
Who it's for
- Agent developers currently using paid APIs (OpenAI, Anthropic, etc.)
- Teams looking to reduce inference costs without sacrificing quality
- Engineers who build with agent frameworks and want cost control
What it won't cover
- How to train or fine-tune models
- Deep learning theory or model architecture details
- Infrastructure setup from scratch
- Enterprise deployment patterns
Skills you'll practise
- Agent workflow cost analysis
- Multi-provider integration
- Performance benchmarking
- Cost-performance optimization decisions