Total Cost
—
Total Tokens
—
Top Project
—
Daily Avg
—
▸ Cost by project
▸ Token breakdown
▸ Daily cost evolution
▸ Projects — detailed breakdown
Loading...
▸ Monthly history
▸ Model pricing reference
How is the cost calculated?
Claude Code charges per token — the unit the AI reads and writes. Every word, symbol, and piece of code is broken into tokens. Your cost depends on how many tokens you sent (input) and how many Claude generated (output).
cost = (input_tokens × $3.00 + output_tokens × $15.00 + cache_read × $0.30 + cache_write × $3.75) ÷ 1,000,000
Cache tokens are cheaper because Claude reuses context it already processed. That's why sessions with lots of repeated code (cache hits) cost much less than fresh conversations.
Example: If your project used 64.3M tokens total (mostly cache reads), it might cost ~$48 — far less than if all were fresh input at $3/M.
| Model | Input / 1M | Output / 1M | Cache read / 1M | Best for |
|---|---|---|---|---|
| Haiku 4.5 | $1.00 | $5.00 | $0.10 | Fast, simple tasks |
| Sonnet 4.6 | $3.00 | $15.00 | $0.30 | Most dev work — best value |
| Opus 4.7 | $5.00 | $25.00 | $0.50 | Complex reasoning, agents |