For your first Prompt optimization task, do not change many settings: run defaults once, then adjust one item at a time. Shorter prompts, less irrelevant context and capped output often cut cost faster than switching models.
Why it matters
Change one variable at a time so you know which setting matters.
Remove irrelevant context.
Step by step
- Measure input, output and retry tokens per user/task.
- Separate cache, batch, embeddings and fine-tuning costs.
- Use 30 days of real requests for cost per user, then margin.
- Model prices change; baseline from official pricing pages and record the date.
Common mistakes
- Estimating from input tokens only; output often costs more.
- Forgetting retries, failures and tool calls.
- Assuming cache hits by default.
- Using stale prices for long-term margin.
Checklist
- Input/output/cache/retry tokens are tracked separately.
- Batch/async fit and latency requirements are clear.
- Embeddings/fine-tuning/image/audio are separate.
- Cost per user and margin use real usage.
- Price date and fallback model are documented.
Key facts
- Remove irrelevant context.
- Cap output.
- Cache repeated parts.
FAQ
How do I start with Prompt optimization?
For your first Prompt optimization task, do not change many settings: run defaults once, then adjust one item at a time. Shorter prompts, less irrelevant context and capped output often cut cost faster than switching models.
How long does it take?
Usually 1-3 minutes; for intermittent issues, run 30 seconds and retest elsewhere.
Is data uploaded?
Runs locally and does not upload video/audio; extensions may still read the page.
Source: Anthropic