cache reads (re-reading context) 21.69B tok $10,843 56%
cache writes (1h) 678M tok $6,781 35%
output (incl. reasoning) 63.5M tok $1,589 8%
fresh uncached input 1.6M tok $8 0%
TOTAL 22.43B tok $19,221
if no caching: $113,418 (5.9x more)
input:output ratio: 353:1
cache hit rate: 97.0%
token_time_breakdown.py (179M unique tokens, 166h wall clock): reasoning (hidden thinking) 29% of tokens, 102h (61%) of time
bash 1.4% 23h (14%)
writing tool calls 4% 14h (8%)
summaries 2.5% 9h (5%)
reading/searching/web 1.6% 7h (4%)
subagents 0.2% 6h (4%)
editing 0.1% 5h (3%)
pasted attachments 25.3% -
typed prompt 34.4% -
system+tools 1.4% -
reread_breakdown.py (per-activity share of billed input): reasoning 59.5% (~$11.4k of the bill is re-sending old
hidden thinking back to the model)
attachments 22.6%
tool calls 7.8%
bash 3.0%
reading/web 2.4%
my prompt 1.6%
summaries 1.5%
system+tools 0.8%
subagents 0.4%
main_vs_sidecar.py: main sidecar combined
sessions/agents 449 484 933
assistant calls 63,820 30,022 93,842
cache hit 97.0% 94.4% 96.6%
turns/agent 142 (median 20, max 11,058 in one session)
62 (median 44)
reasoning % of out 82% 51% 77%
cost @ Opus 4.7 $19,225 $3,495 $22,720
sidecar = 32% of calls but only 15% of cost. Subagents are doing
their job (cheap, focused, short context).
Same shape as yours: re-read dominates, reasoning is the biggest
re-read line, caching is the only thing keeping it sane. The one
that surprised me was a single 11,058-turn main session - some
autonomous loop I forgot to kill. Going to grep for that.
I love PHP since beginning! ZealPHP is my own creation, actively developed. We created it inside our Academy. It powers our EdTech tool today as well. We initially used Apache+mod+php, but as AI needs grew, I don't want to rewrite the whole codebase for asynchronous operations to have SSE, sockets, streaming etc. I used to do with sidecars like RabbitMQ. So it began as an experiment inside the academy, today the same framework powers the EdTech tool we use in-house. To be exact, we made the same Apache+mod_php codebase run in ZealPHP and made it asynchronous.
And why PHP... because it's bloody fast when run asynchronous! HTML rendering is first class, we made ZealPHP yield HTML like react's renderToPipeableStream, and much more using OpenSwoole, making it close to the SAPI contract with classic PHP mental model. So I wanted to explore what PHP could offer in this new direction.
If you like to explore more on how PHP is "bloody fast" and "why": Kindly check https://php.zeal.ninja/performance & https://php.zeal.ninja/why-zealphp
If you like to explore how we made the same codebase run on 2 different server: https://php.zeal.ninja/case-studies/sna-labs
// ZealPHP being Actively developed now!