Costs

Cross-Tenant Token Spend

as of 2026-06-27 14:02 UTC · rolling 30d

Total tokens

5.79B

~$3,284 est.

Input

991.2M

$3.00 / Mtok list

Output

98.3M

$15.00 / Mtok list

Cache read

4.70B

$0.30 / Mtok · 90% off

Est. spend (mo)

~$3.28k

+11.4% vs prior 30d

Daily token volume

Mtok/day · all tenants · 11 buckets

May 28

May 31

Jun 03

Jun 06

Jun 09

Jun 12

Jun 15

Jun 18

Jun 21

Jun 24

Jun 27

Bars in porphyry exceeded the 0.90 Mtok/day soft budget alarm. Spikes correlate with the Jun-18 Ranger fan-out wave and the Jun-27 fleet rehearsal.

Rollup by tenant

WindowModel family

input output cache read · 90% discount

Tenant	Agents	Input	Output	Cache	Total tokens	Share	~$ est.	Trend
Chinchilla (internal) tnt_chinchilla	3	412.6M	38.4M	1.84B	2.29B	100%	~$1,284	stable
Beon IQ — Marta tnt_beon_marta	6	388.1M	41.9M	2.10B	2.53B	94%	~$1,372	rising
Transportation Insights tnt_ti	2	96.3M	9.1M	402.0M	507.4M	24%	~$318	stable
Houston (dev fleet) tnt_houston	5	71.8M	6.7M	288.4M	366.9M	18%	~$236	stable
Sandbox / evals tnt_sandbox	4	22.4M	2.2M	61.0M	85.6M	6%	~$74	stable

Dollar figures are estimates only — derived from public Anthropic list prices ($3 in / $15 out / $0.30 cache-read per Mtok), not billed invoices. Tokens are the metered truth; dollars track them for finance.

Est. spend by tenant

~$ / 30d · list-price model

Chinchilla

~$1,284

Beon IQ — Marta

~$1,372

Transportation Insights

~$318

Houston

~$236

Sandbox / evals

~$74

Margin note

internal · finance

~82%

Blended gross margin across active engagements at current list-price token cost. Cache reads carry ~71% of total volume at a 90% discount, which is the single largest margin lever in the fleet.

Margin = (contracted MRR − est. token cost) ÷ contracted MRR. Excludes infra (EC2, S3, Databricks SP) and is a model, not GAAP. Beon / Marta is the only revenue-bearing tenant today.

static preview — metering pipeline not connected · figures are illustrative