Skip to content

Charts / AI metrics

AI model spend chart

Show spend against the budget, and where the month lands.

Short answer

Show spend to date as a building area split by model, the budget as a line and a forecast to the end of the month. To explain the forecast, add a waterfall from spend to date to the budget.

Updated Nov 2, 2026 · AI sample data

Budget runway: model spend to date $10.2K USDSpend to date is $10,214 after 16 of 31 days; the forecast ends the month at $19,640 (range $19,120 to $20,060) against a $20K budget.Budget runwayQ3 2026MODEL SPEND TO DATE$10.2KUSD$19.6KFORECAST · BUDGET $20K0$10KBUDGET $20K$10,214$19,640OCT 1OCT 8TODAYOCT 23OCT 31SPEND TO DATEATLAS 3ORION XLOTHER MODELSBUDGETSPEND BY MODEL · OCTOBER3 GROUPSAtlas 3$3.9K38%Other models$3.4K33%Orion XL$3.0K29%FORECAST$19.6KOCT 31HIGH CASE$20.1KOVER BUDGETDAILY AVG$63816 DAYSHEADROOM$360AT FORECASTSYNCED 4 MIN AGOOpen report
Model spend against the monthly budget
DaySpend to dateAtlas 3 that dayOrion XL that dayOther models that day
Oct 1$557$211$162$183
Oct 2$1,195$243$185$211
Oct 3$1,843$246$188$213
Oct 4$2,492$247$189$214
Oct 5$3,282$300$229$261
Oct 6$3,645$138$105$120
Oct 7$4,192$208$159$180
Oct 8$4,911$273$209$237
Oct 9$5,643$278$212$241
Oct 10$6,316$256$195$223
Oct 11$7,084$292$223$253
Oct 12$7,773$262$200$228
Oct 13$8,370$227$173$197
Oct 14$8,842$179$137$156
Oct 15$9,514$256$195$222
Oct 16$10,214$266$203$231
  1. 01Spend to date in the heroModel spend so far this month, with the forecast and the budget beside it.
  2. 02Spend by modelThe building area split into the biggest models and the rest.
  3. 03The budget lineThe month's budget, so the forecast reads against it.

The data it needs

ColumnTypeExampleNote
dateDate2026-10-16One row per day and model.
modelTextAtlas 3Group small models into Other.
spend_usdNumber, USD638That day's spend for the model.
budget_usdNumber, USD20000The month's budget, the same on every row.

From spend to date to the budget

From spend to date to the budgetFrom spend to date to the budget: Spend to date $10.2K, Rest of October +$9.4K, Forecast $19.6K, Headroom +$0.4K, Budget $20K.20K15K10K5K0SPEND TO DATE+$9.4KREST OF OCTOBERFORECAST+$0.4KHEADROOMBUDGET
From spend to date to the budget
StepAmount
Spend to date$10.2K
Rest of October+$9.4K
Forecast$19.6K
Headroom+$0.4K
Budget$20K

042 · Waterfall

October has spent $10.2K in 16 days, $638 a day. The rest of the month adds a forecast $9.4K, landing at $19.6K: $360 under the $20K budget, with a high case of $20.1K over it.

The area shows the pace; the waterfall shows how close the month lands. Put it under the widget during the month.

Next to it on an AI dashboard

  • Spend from models through agents to outcomes$10,214 of model spend flows through 6 agents to 4 outcomes; $1,280 ends in retries and failed runs.ATLAS 3$3.9KORION XL$3.0KKITE 2 PRO$2.0KVEGA 1.5$920NOVA$414ANSWERS$7.4KTOOL CALLS$1.5KRETRIES$940FAILED$340TRIAGE$1.6KDOCS$1.2KRESEARCH$1.7KREVIEW$1.0KREFUNDS$1.2KOTHERS$3.5KUSEFULWASTED
    Spend from models through agents to outcomes
    StageNameSpendWasted
    ModelATLAS 3$3,880No
    ModelORION XL$2,960No
    ModelKITE 2 PRO$2,040No
    ModelVEGA 1.5$920No
    ModelNOVA$414No
    AgentTRIAGE$1,640No
    AgentDOCS$1,180No
    AgentRESEARCH$1,730Yes
    AgentREVIEW$1,010No
    AgentREFUNDS$1,190Yes
    AgentOTHERS$3,464No
    OutcomeANSWERS$7,394No
    OutcomeTOOL CALLS$1,540No
    OutcomeRETRIES$940Yes
    OutcomeFAILED$340Yes
    Spend from models through agents to outcomes$10,214 of model spend flows through 6 agents to 4 outcomes; $1,280 ends in retries and failed runs.ATLAS 3$3.9KORION XL$3.0KKITE 2 PRO$2.0KVEGA 1.5$920NOVA$414ANSWERS$7.4KTOOL CALLS$1.5KRETRIES$940FAILED$340TRIAGE$1.6KDOCS$1.2KRESEARCH$1.7KREVIEW$1.0KREFUNDS$1.2KOTHERS$3.5KUSEFULWASTED
    Spend from models through agents to outcomes
    StageNameSpendWasted
    ModelATLAS 3$3,880No
    ModelORION XL$2,960No
    ModelKITE 2 PRO$2,040No
    ModelVEGA 1.5$920No
    ModelNOVA$414No
    AgentTRIAGE$1,640No
    AgentDOCS$1,180No
    AgentRESEARCH$1,730Yes
    AgentREVIEW$1,010No
    AgentREFUNDS$1,190Yes
    AgentOTHERS$3,464No
    OutcomeANSWERS$7,394No
    OutcomeTOOL CALLS$1,540No
    OutcomeRETRIES$940Yes
    OutcomeFAILED$340Yes

    116 · Money flowFlow between steps · AI data

  • Error-budget burn rate per windowError-budget burn rate on a log scale: last hour 6.2×, 6 hours 3.4×, 24 hours 1.3×, this month 0.73×; the last hour pages at 14.4×, 6 hours at 6×, 24 hours opens a ticket at 3×.0.25×1×5×20×LAST HOURPAGES AT 14.4×6.2×6 HOURSPAGES AT 6×3.4×24 HOURSTICKET AT 3×1.3×THIS MONTHUNDER 1×0.73×OVER 1×UNDER 1×PAGING RULE
    Error-budget burn rate per window
    WindowBurn rateRule
    LAST HOUR6.2×PAGES AT 14.4×
    6 HOURS3.4×PAGES AT 6×
    24 HOURS1.3×TICKET AT 3×
    THIS MONTH0.73×UNDER 1×
    Error-budget burn rate per windowError-budget burn rate on a log scale: last hour 6.2×, 6 hours 3.4×, 24 hours 1.3×, this month 0.73×; the last hour pages at 14.4×, 6 hours at 6×, 24 hours opens a ticket at 3×.0.25×1×5×20×LAST HOURPAGES AT 14.4×6.2×6 HOURSPAGES AT 6×3.4×24 HOURSTICKET AT 3×1.3×THIS MONTHUNDER 1×0.73×OVER 1×UNDER 1×PAGING RULE
    Error-budget burn rate per window
    WindowBurn rateRule
    LAST HOUR6.2×PAGES AT 14.4×
    6 HOURS3.4×PAGES AT 6×
    24 HOURS1.3×TICKET AT 3×
    THIS MONTH0.73×UNDER 1×

    114 · Burn-rate tracksCompare items · AI data

  • Model cost against qualityCost per 1K runs against quality for 5 models, from Nova at $4 and 93.2% to Orion XL at $40 and 97.8%; 4 proposed switches, for example Research from Orion XL to Atlas 3 saves $600 a month.1RESEARCH−$6002SALES−$3103REFUNDS+$1904DOCS−$72093%95%97%$10$20$30$40BETTER ↖1234NOVAVEGA 1.5KITE 2 PROATLAS 3ORION XLMODEL · SIZE IS SPENDLATENCY ELEVATEDEFFICIENT EDGE
    Model cost against quality
    ModelCost per 1K runsQualitySpend
    Nova$493.2%$414
    Vega 1.5$1295.6%$920
    Kite 2 Pro$2095.4%$2040
    Atlas 3$2897.4%$3880
    Orion XL$4097.8%$2960
    Switch researchOrion XL → Atlas 3proposed−$600 / mo
    Switch salesKite 2 Pro → Vega 1.5proposed−$310 / mo
    Switch refundsKite 2 Pro → Atlas 3proposed+$190 / mo
    Switch docsAtlas 3 → Vega 1.5off−$720 / mo
    Model cost against qualityCost per 1K runs against quality for 5 models, from Nova at $4 and 93.2% to Orion XL at $40 and 97.8%; 4 proposed switches, for example Research from Orion XL to Atlas 3 saves $600 a month.1RESEARCH−$6002SALES−$3103REFUNDS+$1904DOCS−$72093%95%97%$10$40BETTER ↖1234NOVAVEGA 1.5KITE 2 PROATLAS 3ORION XLMODEL · SIZE IS SPENDLATENCY ELEVATEDEFFICIENT EDGE
    Model cost against quality
    ModelCost per 1K runsQualitySpend
    Nova$493.2%$414
    Vega 1.5$1295.6%$920
    Kite 2 Pro$2095.4%$2040
    Atlas 3$2897.4%$3880
    Orion XL$4097.8%$2960
    Switch researchOrion XL → Atlas 3proposed−$600 / mo
    Switch salesKite 2 Pro → Vega 1.5proposed−$310 / mo
    Switch refundsKite 2 Pro → Atlas 3proposed+$190 / mo
    Switch docsAtlas 3 → Vega 1.5off−$720 / mo

    109 · Frontier mapRelationship · AI data

Common mistakes

Total spend with no budget

Spend alone says nothing about whether it is too much. Draw the budget.

A forecast with no range

Show the high case too, so a near miss is visible.

Every model as its own colour

Past three or four, group the small models into Other.

Get the charts on this page

Budget runway, waterfall and the other 130 types.

In Figma and React, in four sizes, light and dark. Pay once, every update.

101 · Budget runway type page042 · Waterfall type page

See plans

Questions

What counts as model spend?

What the model providers bill for tokens and calls, per model, per day.

How is the forecast made?

From the daily pace so far, with a high case for a busier rest of month. Send it with the data.

Can I draw it from my own data?

Yes. The chart is drawn from your rows (date, model, spend, budget) and lays out again for its size; nothing is scaled.

Does the budget runway come in every size?

Yes: XS, S, M and L, each laid out again for its size, in light and dark.

Charting another metric? Start from what the chart has to show.

Open the Chart Chooser