Jon Marien
current streak0d
requests30,655
spend$3,172.80
tokens7.1B
ONDJFMAMJJASONMWF
spend
$3,172.80
tokens
7.1B tokens
models
23 models
Opus 5$1,248.02 · 39%
Opus 4.7$559.20 · 18%
Opus 4.8$519.09 · 16%
Sonnet 5$311.56 · 10%
Opus 5.5$162.43 · 5%
Fable 5$145.47 · 5%
Gpt 6.1.sol$63.08 · 2%
Gpt 6.astra$40.94 · 1%
what they pay for
$365.00/mo
- AnthropicTeamssubscription$125.00/mo
- OpenAIPlussubscription$20.00/mo
- CursorUltrasubscription$200.00/mo
- WindsurfGrandfather Prosubscription$10.00/mo
- OpenCodeGosubscription$10.00/mo
Self-reported. Nothing here is checked against a provider.
summary
- requests
- 30,655
- active days
- 77
- per active day
- $41.21
- per request
- $0.104
- tokens per request
- 230K
- biggest day
- $224.26
- first reported
- 5 May 2026
- last reported
- 26d ago
tools
4 tools
claude-code$1,768.56 · 56%
claude-desktop$1,221.42 · 38%
codex$160.28 · 5%
opencode$22.54 · 1%
where the money went
| bucket | tokens | cost | share |
|---|---|---|---|
| cache read | 6.8B | $1,769 | 55.8% |
| cache write | 131.8M | $1,081 | 34.1% |
| output | 15.9M | $255 | 8.1% |
| input | 127.9M | $67.54 | 2.1% |
89.8% of this is cache, not completions. Context being re-read each turn, rather than anything typed or generated.
cache reuse
51.5:1
6.8B tokens read back for every 131.8M written. Writing to cache costs 1.25–2× the input rate; reading costs a tenth. Higher is better.
Median is 59.5:1 across 5 people. You are in the 20th percentile. At the median you would have spent about $146 less.
Reuse drops when context is rebuilt: starting fresh sessions, clearing history, or editing an early turn, which invalidates everything cached after it.