What AI APIs actually cost, where the hidden costs appear, and which model is cheapest for the job. Original data, no vendor spin. One email a month.
Read the current edition: AI Agent Cost Index
Join developers and finance teams who plan AI spend with real numbers.
List-price moves across the major API providers, normalized to cost per million tokens.
Retries, long context, reasoning tokens, rate-limit backoff, and idle GPU spend.
Which model is cheapest for coding, support, extraction, and agents — with the trade-offs.
| Input | What we do |
|---|---|
| Public list prices | Normalize input, cached input, and output to cost per 1M tokens. |
| Workload profiles | Model short chat, RAG, coding, and agent loops with retry and context overhead. |
| Measured latency | Report p50/p95 latency separately from price so cheap-but-slow is visible. |
| Change log | Show what moved since the last issue, not just today's snapshot. |
Is it free? Yes. No Plugsky account required.
How often? Once a month. One-click unsubscribe in every issue.
Do you sell my email? No. It is used only for the AI Cost Index and occasional product updates.