bosslegend

@boss

Published API price

$6,088

What these tokens would have cost billed per token at each model's published price. It is the one figure that compares across people on different plans.

API list price / day · last 90 days

on a plan$5,333

paid per token$755.48

per active day$62.77

10.97B

tokens

113.1M on a working day

97,897

requests

112.1K each

active days

not a day missed

day streak

still running

Activity

+63% vs previous 30 days

10.97B tokens over 97 active days · last 12 months

Averaging 113.1M per working day, with a median of 37.9M. Busiest was Sep 14, 2026 at 802.2M. Cells are shaded by the day's published API price. Hover a day for the models and tools behind it.

Last 12 months in full on this profile, last synced Sep 24, 2026. Counters only: no prompts, no project names, no file paths.

Drift

What got reached for, month by month

Each row is one month scaled to its own 100%, so a quiet month and a loud one take the same width and only the mix changes.

Jun 26

Jul 26

Aug 26

Sep 26

muse-spark-1.3claude-sonnet-5claude-opus-5muse-spark-1.3-contributormuse-spark-1.3-contributor-freeclaude-fable-5claude-opus-4-8claude-haiku-4-5other

Where it goes

$6,088, split four ways

Output is a fraction of the tokens but usually most of the money: it is priced at several times the rate of cache reads.

cache write

$2,501

cache read

$2,375

output

$929.82

fresh input

$282.05

Every model here has a known rate.

Cache

Cache does most of the reading

A raw token total mostly measures re-read context. The ratios below show how the work was done.

96.2%

of everything entering the context window came from cache instead of being paid for again

114.7×

tokens of context read per token written

112.1K

tokens in an average request

cache read

91% · 9.97B

cache write

5% · 518.2M

output

1% · 94.8M

fresh input

4% · 392.7M

Models

33 models, mostly muse-spark-1.3

muse-spark-1.3

3.15B · 28.8%

$745.60 estimated at API list price · 24,490 requests

claude-sonnet-5

2.68B · 24.4%

$1,223 estimated at API list price · 21,325 requests

claude-opus-5

2.37B · 21.6%

$2,598 estimated at API list price · 15,614 requests

muse-spark-1.3-contributor

1.16B · 10.6%

$9.52 estimated at API list price · 9,927 requests

muse-spark-1.3-contributor-free

491.8M · 4.5%

$16.21 estimated at API list price · 7,304 requests

claude-fable-5-1

224.0M · 2%

$397.39 estimated at API list price · 759 requests

claude-fable-5

185.0M · 1.7%

$535.58 estimated at API list price · 2,124 requests

gemini-3.8-flash

177.2M · 1.6%

$24.03 estimated at API list price · 1,084 requests

claude-opus-4-8

131.5M · 1.2%

$220.81 estimated at API list price · 1,448 requests

claude-haiku-4-5

123.3M · 1.1%

$66.23 estimated at API list price · 4,240 requests

+23 more below one percent.

Tools

5 tools, mostly Claude Code

Claude Code

5.84B · 52,046 requests · $5,246

53.2%

Muse

4.26B · 33,137 requests · $738.84

38.9%

OpenCode

552.0M · 8,452 requests · $29.02

Antigravity

303.5M · 3,999 requests · $70.93

2.8%

Codex

12.5M · 263 requests · $3.89

0.1%

Rhythm

Heaviest on Monday

Tokens by day of the week, in the timezone each device reported.

Mon: 2.23B over 14 days

Tue: 2.03B over 14 days

Wed: 1.63B over 14 days

Thu: 2.10B over 14 days

Fri: 1.40B over 13 days

Sat: 803.3M over 14 days

Sun: 782.8M over 14 days

By month

4 months tracked, busiest Sep 26.

Jun 26: 16.9M over 11 days

Jul 26: 889.7M over 31 days

Aug 26: 4.17B over 31 days

Sep 26: 5.89B over 24 days

Hour of day is not in the sync payload yet, so the rhythm here stops at the day. See docs/tokenstat-sync-cli.md for what the CLI would need to send.

Spread

Across 1791 projects

Project names stay on the device. What reaches us is an opaque hash we can count but never read, so this page shows how the work was spread, never which work it was.

project 1

30.6%

project 2

29.9%

project 3

15.9%

project 4

12.2%

project 5

4.5%

project 6

4.4%

project 7

0.9%

project 8

0.7%

Ordered by size, unlabelled on purpose. The top project is 30.6% of everything.

How the counts were read

Some tools report token counts exactly; the rest are derived from what they write to disk. The chart keeps that split visible, so an inferred number never poses as a measured one.

reported by the coding tool

98.6%

reconstructed, high confidence

1.2%

derived from local files

0.2%