Build a Local Model Eval Harness

A local model course needs an eval harness so students can compare families, quantizations, prompts, and runtimes with evidence. In local AI, the model family is only one part of the system. The runtime, file format, serving path, hardware budget, evaluation set, and safety policy decide whether the model becomes useful.

Current source signal

Build the small version

Create a 25-case eval set with categories for chat, code, RAG, JSON, safety, and speed.

1. 1Define the user task in one sentence.
2. 2Choose the smallest model and runtime that might pass that task.
3. 3Run one happy-path prompt and one failure-path prompt.
4. 4Record speed, memory pressure, output quality, and the exact reason for any failure.
5. 5Write the operating rule you would give a non-expert user.

eval_harness:
  cases:
    - id
    - category
    - prompt
    - expected_behavior
    - scoring_rubric
  run_against:
    - model_name
    - quantization
    - runtime
  output:
    - score
    - latency
    - failure_notes

The big idea: evidence beats demos. A local model app is not done when the model answers once; it is done when the whole workflow can be installed, measured, trusted, and recovered.