evals
Create your eval

03 — Investigation · Is it worth the cost?

Is the expensive model worth it?

32 runsRound 13 game months per turnrunbook v1.0.0collection jettygrowthteamtask micropolis-cityupdated 2026-09-28

Decision it informs: Which model to pay for when a long-horizon agent task runs many times.

Key takeaway generated

GPT-5.6 Terra met goals for the least money ($7.55 per goal met); per-run cost ranged from $0.08 (GPT-5.6 Luna) to $1.76 (Jev Router).

At a glance computed

Runs
32
one city per Jetty run
Model spend
$28.08
OpenRouter, the mayor models
Goals met
1/32
judge score ≥ threshold
Eval pass rate
3%
1/32 runs valid and goal met

Key figure computed

Cost per run vs share of goals met

One dot per model over all its runs. Up and to the left is better.

Table view
ModelCost / runGoals metCost / goal metLatencyGrow pop.
GPT-5.6 Terra$0.941/8$7.5515s3,290
Jev Router$1.760/8–48s12,120
GPT-5.6 Luna$0.080/8–7s7,210
Claude Sonnet 5$0.720/8–10s1,070

Evaluations computed

Every run's evals-v2 validation report, scored by Jetty: code checks (is the run valid), the judge (was the goal met) and the reviewer checklist.

EvaluationPassedRateShare
Code checks (run is valid)192/192100%
Checklist items96/96100%
Judge: task goal met1/323%
Valid runs (all code checks and checklist items)32/32100%
Runs with a passing verdict (valid and goal met)1/323%

Findings generated

Cost per run spans 22×

high confidence

Mean model spend per run: GPT-5.6 Luna $0.08; Claude Sonnet 5 $0.72; GPT-5.6 Terra $0.94; Jev Router $1.76.

32 runsmodel_cost_usd in runs.csv

GPT-5.6 Terra buys a met goal for the least money

medium confidence

Model spend per goal met: GPT-5.6 Terra $7.55; Jev Router no goals met; GPT-5.6 Luna no goals met; Claude Sonnet 5 no goals met.

GPT-5.6 Terra 1/8

Latency, not game time, is what slow models pay

high confidence

Mean seconds per turn: GPT-5.6 Luna 7s; Claude Sonnet 5 10s; GPT-5.6 Terra 15s; Jev Router 48s. The game waits for every answer, but the 45-minute budget does not: 3 runs handed their last turns to autopilot.

Jev Router 3 autopilot runs

Proposed improvements generated

Runbook

Trim the text map

The 60×50 map is most of each prompt; sending only changed rows after turn 1 would cut prompt cost for every model.

backed by: briefings in decisions.jsonl

Run

Try the cheapest model with more turns

Shorter turns (1 month) cost the cheap model little and may close the gap with the expensive one.

backed by: GPT-5.6 Luna $0.08 a run

Caveats

  • Costs are OpenRouter charges for the mayor model only, not the runbook agent or the sandbox.
  • Router pricing varies per request.
  • One or two runs per model and task (see the matrix).
Runs and provenance
RunTaskSeedMayorOutcomePopulationCity scoreEvaluationModel costWall minJetty run ID
jev-grow-s2Grow a city2Jev Router19,660 people (bankrupt)19,660596fail$2.1446.35de24fe4
sonnet5-grow-s2Grow a city2Claude Sonnet 51,200 people1,200431fail$1.3914.240c8cf3e
gpt56terra-grow-s2Grow a city2GPT-5.6 Terra3,660 people3,660322fail$1.5413.0476d94f4
gpt56luna-grow-s2Grow a city2GPT-5.6 Luna8,060 people8,060417fail$0.148.4831ede26
jev-grow-s6Grow a city6Jev Router4,580 people4,580367fail$1.9146.6981094de
sonnet5-grow-s6Grow a city6Claude Sonnet 5940 people940514fail$1.3213.10e6f2bec
gpt56terra-grow-s6Grow a city6GPT-5.6 Terra2,920 people (bankrupt)2,920756fail$1.4112.64c859227
gpt56luna-grow-s6Grow a city6GPT-5.6 Luna6,360 people6,360311fail$0.159.7349eb345
jev-tokyo-s1Tokyo 19571Jev Routerlost23,120372fail$1.2114.496f24e6b
sonnet5-tokyo-s1Tokyo 19571Claude Sonnet 5lost23,800333fail$0.393.458cccc15
gpt56terra-tokyo-s1Tokyo 19571GPT-5.6 Terrawon29,640501pass$0.616.000ff7869
gpt56luna-tokyo-s1Tokyo 19571GPT-5.6 Lunalost23,700330fail$0.042.54675adce
jev-tokyo-s2Tokyo 19572Jev Routerlost21,900368fail$0.6510.1d78fe456
sonnet5-tokyo-s2Tokyo 19572Claude Sonnet 5lost22,400346fail$0.383.172ed89e1
gpt56terra-tokyo-s2Tokyo 19572GPT-5.6 Terralost31,020237fail$0.555.42cf5d9c4
gpt56luna-tokyo-s2Tokyo 19572GPT-5.6 Lunalost29,120458fail$0.042.5a57b0db6
jev-sanfrancisco-s1San Francisco 19061Jev Routerlost36,300286fail$0.8610.7b6be516e
sonnet5-sanfrancisco-s1San Francisco 19061Claude Sonnet 5lost36,84023fail$0.423.7df629564
gpt56terra-sanfrancisco-s1San Francisco 19061GPT-5.6 Terralost47,68039fail$0.656.4d5eb8764
gpt56luna-sanfrancisco-s1San Francisco 19061GPT-5.6 Lunalost23,68021fail$0.052.67a99b8d9
jev-sanfrancisco-s2San Francisco 19062Jev Routerlost92,360500fail$2.0018.4d76eb040
sonnet5-sanfrancisco-s2San Francisco 19062Claude Sonnet 5lost72,30027fail$0.454.3fd728217
gpt56terra-sanfrancisco-s2San Francisco 19062GPT-5.6 Terralost38,88037fail$0.686.4b8e5bc93
gpt56luna-sanfrancisco-s2San Francisco 19062GPT-5.6 Lunalost48,22022fail$0.042.5cd12e212
jev-detroit-s1Detroit 19721Jev Routerlost7,380595fail$2.2126.2aea16d25
sonnet5-detroit-s1Detroit 19721Claude Sonnet 5lost3,520586fail$0.736.9650b3bc0
gpt56terra-detroit-s1Detroit 19721GPT-5.6 Terralost6,120597fail$1.069.8175f2d53
gpt56luna-detroit-s1Detroit 19721GPT-5.6 Lunalost7,400547fail$0.084.8fde1c078
jev-detroit-s2Detroit 19722Jev Routerlost10,040456fail$3.1145.985fe296a
sonnet5-detroit-s2Detroit 19722Claude Sonnet 5lost3,280341fail$0.726.6c55e7a2b
gpt56terra-detroit-s2Detroit 19722GPT-5.6 Terralost6,880693fail$1.0510.8ef93e671
gpt56luna-detroit-s2Detroit 19722GPT-5.6 Lunalost15,280704fail$0.095.4aba6000b

Ready to stop guessing if your outputs are good?

Create your eval Book a 15-minute walkthrough
Connect your agent to Jetty →