Skip to main content
Skip to content
Menu

Open Source Beginner 3 parts

Select the right open model size for your specific reasoning task

You'll have documented which open model tier (2B through 31B parameters) handles your particular workflow best, including performance benchmarks you measured yourself and cost-per-inference data for y

What you'll make

Open Model Selection Brief

  1. Task Performance Benchmarks: measured accuracy and speed across model tiers
  2. Resource Requirements: hardware compatibility and cost-per-inference analysis
  3. Selection Decision: chosen model with supporting evidence and deployment constraints

The 3 parts

Each part ends with a check you can see.

  1. Test Models on Your Task

    Run the same reasoning task across multiple model sizes to measure actual performance differences

    Done when

    You have performance data (accuracy, response time) for at least 3 model sizes on your specific task, with at least one smaller model that outperforms expectations

  2. Map Hardware and Cost Constraints

    Document inference costs and hardware requirements to understand the true resource impact of each model option

    Done when

    You have cost-per-inference data for each tested model on your target hardware, plus minimum system requirements documented for deployment decisions

  3. Choose Your Model with Evidence

    Make and defend your model selection decision using performance and cost data from your testing

    Done when

    You have a documented model selection with supporting performance data, cost justification, and deployment plan that accounts for your specific constraints

    Capstone: you put together your Open Model Selection Brief.

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

  • Select optimal open model size for your specific reasoning workload using measured performance data
  • Benchmark inference costs across model tiers to make defensible resource allocation decisions
  • Deploy and test models efficiently on your actual hardware constraints
  • Build a reusable evaluation framework for future model selection decisions

Who it's for

  • Developers and technical leads choosing between open models for production workloads
  • Teams with compute budget constraints who need maximum reasoning per dollar
  • Organizations evaluating local model deployment vs API costs

What it won't cover

  • Model training or fine-tuning techniques
  • Prompt engineering optimization
  • Building custom model architectures
  • Enterprise security and compliance considerations

Skills you'll practise

  • Performance benchmarking across model tiers
  • Resource constraint analysis
  • Cost-per-inference calculation
  • Hardware-model compatibility assessment