What AI model should I use for lead qualification?

Use a two-stage approach for cost/accuracy balance.

Model Comparison

Model	Cost/1K tokens	Accuracy	Best For
GPT-5-mini	$0.00015	70-80%	Initial filtering, high volume
GPT-5	$0.005	90-95%	Final qualification, personalization
Claude 3 Haiku	$0.00025	75-85%	Balanced cost/accuracy
Claude 3.5 Sonnet	$0.003	92-97%	Complex qualification

Recommended Strategy

Stage 1 - Fast Filter (High Volume)

• Model: GPT-5-mini
• Batch size: 100
• Goal: Quick yes/no filtering

Stage 2 - Deep Qualification (Borderline Cases)

• Model: GPT-5
• Batch size: 25
• Filter: Only re-check where stage1_score >= 0.5

Stage 3 - Personalization (Qualified Leads)

• Model: GPT-5
• Batch size: 10
• Goal: First line generation, custom messaging

Key Insight

It might be worth spending more money on a more expensive model than spending countless hours tweaking the prompt.

GPT-5 often gets the right answer immediately where GPT-5-mini fails.