This is a planning estimate, not a quote. Real bills depend on prompt size, context length, retries, and how each provider meters tokens on any given day.
- A working month is assumed to be 22 days. "Light / Moderate / Heavy" maps to roughly 15 / 40 / 100 AI interactions per coding day.
- Inline code completions are treated as free where the tool includes them (Copilot does), so only chat and agent work is metered.
- Agent-heavy work costs more per action because one agent run fires several model calls. Frontier models cost several times more per action than budget ones.
- Figures are shown as a range of about plus or minus 25 percent to reflect that spread.
- Team estimates multiply individual-plan rates by seats. Real team and business tiers add admin controls at a somewhat higher per-seat price, so treat multi-seat totals as a floor.