Model Test Harness
Money Now Matrix · Quick-Add eval benche.g. "send invoice acme 15m due tue" · "30 min yoga tomorrow"
Context & sampling ›
When it matters
Temperature 0.20
prod = 0.20
Default duration when none is typed (min)
Sent to the worker, which fills it in when no duration is found in the task. The worker no longer rejects duration-less tasks.
Role / business context
North-star goals (one per line)
System prompt ›
Prefilled with the production MNM prompt. Edit to A/B a prompt change across models.
Model comparison bench
Type a task, check the models you want to compare, and hit Run. Each model gets the same query and the same MNM system prompt; you'll see the Priority Category, reasoning, scores, latency, tokens, and cold-vs-warm cost side by side.