What these tokens would have cost billed per token at each model's published price. Not a bill. It is the one figure that compares across people on different plans.
List-rate estimate / day
Jun 17, 2026peak $477.31Aug 8, 2026
on a plan$9,187
paid per token$158.11
per active day$176.32
12.41B
tokens
234.2M on a working day
69,111
requests
179.6K each
53
active days
not a day missed
53
day streak
still running
Activity
12.41B tokens over 53 active days · last 12 months
Averaging 234.2M on a day worked, with a median of 190.3M. Busiest was Aug 6, 2026 at 1.07B. Cells are shaded by what the day was worth at list rates, not by token count. Hover a day for the models and harnesses behind it.
less spendmore spend
Last 12 months in full on this profile, last synced Aug 8, 2026. Counters only: no prompts, no project names, no file paths.
Drift
What got reached for, month by month
Each row is one month at 100% of its own tokens, so a quiet month reads the same width as a loud one and the only thing changing is the mix.
Output is a fraction of the tokens and usually most of the money, because it is priced at several times the rate of anything read back out of cache.
cache read
$4,742
output
$2,826
cache write
$1,583
fresh input
$194.01
Every model in this archive has a known rate. 2.2% of them ran on models whose id we can only pattern-match, so their rate is our reading of a public price rather than an exact match.
Cache
Reuse, not volume
A raw token total mostly measures how much context got re-read. The ratios underneath it are the part that says anything about how the work was done.
99.1%
of everything entering the context window came from cache instead of being paid for again
115.3×
tokens of context read per token written
179.6K
tokens in an average request
cache read
97% · 11.98B
cache write
2% · 217.3M
output
1% · 106.7M
fresh input
1% · 109.4M
Models
37 models, mostly claude-opus-5
claude-opus-5
2.77B · 22.3%
$1,925 estimated at list rates · 11,480 requests
deepseek-v4-flash
2.41B · 19.4%
$2.95 estimated at list rates · 5,213 requests
claude-opus-4-8
2.38B · 19.2%
$2,580 estimated at list rates · 15,473 requests
claude-fable-5
1.64B · 13.2%
$3,548 estimated at list rates · 9,435 requests
claude-sonnet-5
1.17B · 9.4%
$458.26 estimated at list rates · 10,343 requests
cursor-grok-4.5-high-fast
550.4M · 4.4%
$253.13 estimated at list rates · 862 requests
grok-4.5
328.6M · 2.6%
$127.14 estimated at list rates · 2,201 requests
cursor-router-auto
275.9M · 2.2%
$26.03 estimated at list rates · 390 requests
gpt-5.6-luna
266.3M · 2.1%
$34.12 estimated at list rates · 1,549 requests
claude-sonnet-4-6
182.8M · 1.5%
$156.25 estimated at list rates · 2,717 requests
+27 more below one percent.
Tools
10 harnesses, one archive
Claude Code
8.30B · 57,210 requests · $7,093
66.9%
Codex
2.34B · 4,923 requests · $15.64
18.8%
Cursor
914.2M · 1,352 requests · $319.74
7.4%
OpenCode
346.9M · 2,044 requests · $27.27
2.8%
Grok Build
328.7M · 2,208 requests · $127.20
2.6%
Antigravity
121.9M · 1,191 requests · $50.22
1%
Claude Code rollup
58.7M · 92 requests · $1,711
0.5%
Cline
386.7K · 79 requests · $0.14
0%
OpenClaw
230.1K · 11 requests · $0.05
0%
Copilot CLI
17.0K · 1 requests
0%
Rhythm
Heaviest on Thursday
Tokens by day of the week, in the timezone each machine reported.
Mon: 1.65B over 7 days
Tue: 1.58B over 7 days
Wed: 1.89B over 8 days
Thu: 2.71B over 8 days
Fri: 1.99B over 8 days
Sat: 1.39B over 8 days
Sun: 1.19B over 7 days
MTWTFSS
By month
3 months tracked, busiest Jul 26.
Jun 26: 750.3M over 14 days
Jul 26: 7.62B over 31 days
Aug 26: 4.04B over 8 days
JunJulAug
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 70 projects
Project names stay on the machine. What reaches us is an opaque hash we can count but never read, so a profile can show how the work was spread and never which work it was.
project 1
24.1%
project 2
19.4%
project 3
19.1%
project 4
12%
project 5
7.4%
project 6
7.4%
project 7
6.1%
project 8
1.5%
Ordered by size, unlabelled on purpose. The top project is 24.1% of everything.
How the counts were read
Some harnesses report token counts exactly; others have to be derived from what they write to disk. The archive keeps that distinction rather than flattening it, so a number that was inferred never passes itself off as a measurement.
