New Evaluation

Select task types, models, and test prompts. Multi-select supported across all sections.

1. Task Type

1 selectedMulti-select

Choose the category that best matches what you need the LLM to do. Select multiple types to get merged criteria and compare models across task families.

Cold Email
GTM

Sales, marketing, and revenue tasks — cold outreach, lead scoring, proposals, and content.

General

Broad language tasks — coding, summarisation, reasoning, and general knowledge.

🏦FinServ• FinServ Pack scored

Regulated financial services tasks. Automatically adds hallucination, compliance, and auditability scoring for high/critical risk.

2. Select Models

🔍

3. Test Prompts

Tips for good eval prompts
  • Write 3–5 prompts that represent real tasks you’d run in production — not edge cases.
  • Be specific: include context like industry, company size, tone, or output format where relevant.
  • Vary difficulty: include at least one straightforward and one more nuanced prompt.
  • The AI judge scores each response independently — more prompts = more reliable WyzScore.
One prompt per row · first column used · max 20 prompts
Cancel