TOKEN ECONOMY

What the tokens cost, and where they went

Every connected account, by model and by day — input, cached and output tokens against the price list, plus what none of it could account for.

Open assessments

2026-08-01 → 2026-08-30· 30 days360 readings · 4 of 6 accounts · 9 models

Model cost

$101,200

18%vs previous period

LOGGEDover the window

Total tokens

64.49B

16%vs previous period

LOGGEDinput + output + reasoning

Cached input

23%

6.0%vs previous period

LOGGED13.64B tokens billed at a tenth

Cost per 1M tokens

$1.57

1.2%vs previous period

MODELLED$0.0016 per 1K

Models in use

of 10 priced · 3 of 16 providers

Accounts reading

of 6 connected · 2 need you

Waiting on you

2 accounts cannot proceed. $12,250/month is on the bill and unreadable in the logs.

Resolve

Every figure states how it was obtained:MEASUREDLOGGEDMODELLED

Token economy

Where the tokens go, and what each one costs

gemini-2.5-pro$32,16318.28B

$0.0018/1K · $1.76/1M · 22% cached

gpt-4.1$23,74510.44B

$0.0023/1K · $2.27/1M · 6% cached

claude-sonnet-4-5$18,1525.74B

$0.0032/1K · $3.16/1M · 32% cached

gpt-4.1-ft-vega01$17,8005.11B

$0.0035/1K · $3.49/1M · 2% cached

gpt-4.1-mini$3,1407.29B

$0.0004/1K · $0.43/1M · 14% cached

gemini-2.5-flash$2,7657.81B

$0.0004/1K · $0.35/1M · 32% cached

claude-haiku-4-5$2,7593.26B

$0.0008/1K · $0.85/1M · 45% cached

gemini-2.5-flash-lite$3624.54B

$0.0001/1K · $0.08/1M · 53% cached

gpt-4o-mini$3152.03B

$0.0002/1K · $0.16/1M · 33% cached

Fresh input72%· full rateCached input21%· a tenth of the rateOutput6%· the dearest tokensReasoning1%· billed as output

Token usage by day

Same data, four ways to cut it

gemini-2.5-progpt-4.1claude-sonnet-4-5gpt-4.1-ft-vega01gpt-4.1-minigemini-2.5-flashclaude-haiku-4-5gemini-2.5-flash-litegpt-4o-mini

Cache effectiveness

Cached share against cost per million — bubble size is volume

LOGGED

Caching is the one lever that lowers cost without touching the model or the output. A model sitting high and left is paying full rate on input it has seen before.

gemini-2.5-progpt-4.1claude-sonnet-4-5gpt-4.1-ft-vega01gpt-4.1-minigemini-2.5-flashclaude-haiku-4-5gemini-2.5-flash-litegpt-4o-mini

Evidence behind these figures

How the $101,200 above was arrived at

LOGGED$73,63373%

token counts read straight from your provider's logs — the bulk of any bill

MODELLED$24,83625%

computed where coverage left a gap, from the rates beside it

MEASURED$2,7313%

the sampled traffic a completed run actually replayed and observed

These three account for the whole window. Nothing here is assumed — every figure on this page was either read, replayed, or computed from something that was.

What is missing · monthly rate

Unattributed$33,125: Spend inside accounts we are connected to, on calls whose logs did not say which application made them. We know the amount; we do not know the owner. It is what 72% coverage leaves behind — closing it means better log labelling at the source, not another connector.
Unread$46,950: Spend in accounts nothing is connected to. Enumerated and visibly busy, but no connection reads them, so none of it reaches any figure on this page. Connecting them is the fix.

Neither is folded into the totals above, and both are monthly rates — they do not narrow when you shorten the window.

Usage detail

Latest day (2026-08-30) above, whole window below

Model	RequestsLatest / period	InputLatest / period	CachedLatest / period	OutputLatest / period	TotalLatest / period	CostLatest / period	Rateper 1M
gemini-2.5-pro Google Vertex AI	293.1k6.8M	888.5M16.82B	157.6M3.73B	52.8M1.01B	961.2M18.28B	$1686.34$32,163	$1.76
gpt-4.1 Azure AI Foundry	139.4k4.1M	343.1M9.83B	16.3M557.6M	20.1M618.4M	363.3M10.44B	$821.47$23,745	$2.27
claude-sonnet-4-5 Anthropic	94.4k2.1M	174.5M5.28B	51.4M1.71B	9.7M322.5M	189.7M5.74B	$610.44$18,152	$3.16
gpt-4.1-ft-vega01 Azure AI Foundry	48.9k1.9M	161.7M4.80B	3.7M117.8M	8.4M304.9M	170.1M5.11B	$577.86$17,800	$3.49
gpt-4.1-mini Azure AI Foundry	87.8k2.9M	207.7M6.86B	29.6M937.8M	11.6M422.3M	219.3M7.29B	$92.01$3,140	$0.43
gemini-2.5-flash Google Vertex AI	94.7k3.1M	235.2M7.39B	59.7M2.34B	15.6M425.2M	250.8M7.81B	$97.37$2,765	$0.35
claude-haiku-4-5 Anthropic	28.2k1.2M	72.5M3.07B	31.5M1.38B	4.3M187.3M	76.9M3.26B	$63.35$2,759	$0.85
gemini-2.5-flash-lite Google Vertex AI	84.0k1.8M	144.2M4.27B	86.7M2.24B	10.9M267.9M	155.1M4.54B	$12.26$362	$0.08
gpt-4o-mini Azure AI Foundry	32.1k792.8k	59.8M1.90B	20.8M625.2M	3.2M122.4M	63.0M2.03B	$11.25$315	$0.16

Cost is derived from token counts against the published price list, so the rate in the last column reconciles with the totals beside it rather than being an average of averages.