Model Fit Test

Your task, your rules. Run it on the new model. See if it fits you.

A new model is out and everyone says it is great. Whether it is good enough depends on your work. This page builds a short test from your own tasks. You run it in your own chat app and read the verdict. It works for any model.

This is a quick check on your own tasks, not a benchmark. See how a model handles your real work first, then decide if it is worth rebuilding any workflows around.

Describe your work

Pick the closest niche to load three example tasks, then replace them with real ones. Real tasks give a real answer.

Niche

These are example tasks. Edit them or paste your own.

Run the test

  1. Open Claude and start a new chat.
  2. In the model picker, make sure the model you want to test is selected.
  3. Paste this prompt and send it.
  4. Read the stamp at the end of the reply. A model grading itself leans generous, so treat a FIT as promising. Switch to Strict for a harder check.
Prompt · test and self-grade

Read the stamp

Each task gets a 1 to 5 score, and the result ends with one of three verdicts.

FIT
Ready to train

Switch these tasks to the new model and keep the savings.

Average 4 or more, no task under 3
BORDERLINE
Needs a warm-up

Good for some tasks. Route the weak ones to a bigger model.

Average 3 to 3.9, or one task under 3
NOT YET
Rest day

Stay where you are. Test again after the next update.

Anything lower