# Token Budget Calculator

> AI Starter Package (aistarterpackage.com)

## Model Pricing (March 2026)

| Model | Input (per 1M tokens) | Output (per 1M tokens) | Best For |
|-------|----------------------|----------------------|----------|
| Claude Haiku 4.5 | $0.80 | $4.00 | Formatting, boilerplate, simple tasks |
| Claude Sonnet 4.6 | $3.00 | $15.00 | Code review, moderate complexity |
| Claude Opus 4.6 | $15.00 | $75.00 | Architecture, complex reasoning |
| GPT-4o | $2.50 | $10.00 | General purpose, vision |
| GPT-4o mini | $0.15 | $0.60 | Simple tasks, high volume |
| Gemini 2.5 Pro | $1.25 | $10.00 | Long context, multimodal |

## Monthly Budget Estimator

### Light Usage (solo developer, 2-3 hours/day)
- ~50K input tokens/day × 30 days = 1.5M input/month
- ~20K output tokens/day × 30 days = 600K output/month
- **With Sonnet:** ~$13.50/month
- **With Opus:** ~$67.50/month

### Medium Usage (full-time AI-assisted development)
- ~200K input tokens/day × 30 days = 6M input/month
- ~80K output tokens/day × 30 days = 2.4M output/month
- **With Sonnet:** ~$54/month
- **With Opus:** ~$270/month

### Heavy Usage (agent swarms, automation)
- ~500K input tokens/day × 30 days = 15M input/month
- ~200K output tokens/day × 30 days = 6M output/month
- **With Sonnet:** ~$135/month
- **With Opus:** ~$675/month

## Cost Reduction Strategies

1. **Model Routing (saves 30-40%):** Use Haiku for simple tasks, Sonnet for medium, Opus only for complex
2. **Context Pruning (saves 20-30%):** Read only files you need, keep CLAUDE.md under 100 lines
3. **Caching (saves 10-20%):** Anthropic prompt caching reduces repeat context costs by 90%
4. **Batch Processing (saves 15%):** Group similar tasks instead of one-off requests

## Quick Formula

```
Monthly Cost = (Input Tokens × Input Price) + (Output Tokens × Output Price)

Example: 3M input + 1M output on Sonnet
= (3 × $3.00) + (1 × $15.00) = $9 + $15 = $24/month
```

---

*Get the complete token optimization guide at aistarterpackage.com/guides/token-optimization*