Choosing a Model

Pick the right AI model for your use case and budget.

Available Models

Decision Guide

Use DeepSeek (default) when:

• You want the best balance of cost and quality
• Your document set is under 164K tokens
• Standard analysis and summarisation tasks

uv run python analyze.py results.json \
  --context "Summarise key findings" \
  --model deepseek

Use Gemini Flash when:

• You have a very large document set (>500 documents)
• You need the 1M token context window
• Cost sensitivity is moderate

uv run python analyze.py results.json \
  --context "Analyse all documents" \
  --model gemini-flash

Use Ollama (local) when:

• Privacy is the top priority — nothing leaves your machine
• You want free analysis with no API costs
• You accept lower quality compared to cloud models

uv run python analyze.py results.json \
  --context "Summarise findings" \
  --local

Use the Full Pipeline when:

• You want to minimise costs on large document sets
• Free local triage filters to high-relevance documents first
• Only high-scoring documents go to the paid model

uv run python analyze.py results.json \
  --full-pipeline \
  --context "Identify key themes" \
  --model deepseek

Typical Costs (DeepSeek V3.2)

Cost Safety

All paid API calls show a cost estimate and require y/n confirmation before proceeding. Configure limits in config.yaml:

analysis:
  confirm_before_api_call: true
  max_cost_per_query: 1.00
  warn_above: 0.10