MOOTER · METHODOLOGY · DES. 002ESC 1:1 · REV v1.53.0 · 2026-08-27

Show me my number.

Plug in your actual setup — hardware, OS, subscriptions, usage. Mooter projects your tier mix and monthly cost against the all-Opus baseline, computed live from the same per-tier costs as the benchmark below. Every routing decision is logged so you can verify it yourself.

calculatorscenarioprofile assumed by GPU class — not a measurement
Step 1 · Hardware
Step 2 · Operating system
Step 3 · Subscriptions
Step 4 · Usage pattern
Your monthly projection
without mooter
$100.80
all-Opus on every prompt
with mooter
$9.47
mooter-routed
difference
$91.33
in this scenario
Hypothetical scenario, not a measurement. The tier distribution above is an assumed profile per GPU class, and the per-prompt costs are token estimates — nothing on this panel was measured on your machine. We publish no savings percentage, here or anywhere — see the honest numbers. What is measured, on your own machine and by you: mooter recibo reads the token counts Claude Code already writes to your transcripts, attributes every API call to the prompt that caused it, and prices it — no estimate anywhere in that number.
Predicted tier distribution
T0local · Ollama
61%
T1haiku · gpt-4o-mini
18%
T2sonnet · gpt-4o
14%
T3opus · o1-pro
8%
Your stack supports
All local models (24+ GB GPU)
Sonnet & Opus (Anthropic)
GPT-4o (OpenAI)
one case~80 prompts/dayassumed profile, not a herd average

A concrete case

Solo founder, Claude Code Max plan, RTX 4090, ~80 prompts/day, ~8% critical. Most of the day is renames, commits, small edits and “explain this” — those route to T0 local (free). The ~8% that's real debugging or a cross-file refactor goes to Sonnet/Opus. Set those inputs above to see this profile's monthly figure — it's computed from the same per-tier costs as the N=34 benchmark below, not a marketing number. Your mix shifts with how much you keep local.

benchmarkN=343 arms · blind judge

Benchmark proof

Cost per prompt on a 34-prompt blind-judged validation set.

Mooter
$0.022
Sonnet-only
$0.028
Opus-only
$0.034
reproduce34 × 3pre-registered · open repo

Reproduce it yourself

The benchmark is pre-registered and open. 34 prompts × 3 arms (mooter, Sonnet-only, Opus-only) with a blind LLM judge — design, prompts, raw rows and per-pack diagnostics all live in the repo. Clone it and run the harness:

git clone https://github.com/pauloloureiroshp-ship-it/mooter cd mooter/packages/router/scripts/wave1-benchmark tsx run.ts # full run: 34 × 3 arms → outputs/

See wave1-benchmark/README.md + BENCHMARK_DESIGN.md for the full method, confidence intervals and mis-routing analysis. Pinned to mooter v1.53.0.

N=34 is a small set — only medium-to-large effects are detectable. On this cloud-only set mooter matches the quality bar at lower cost per prompt; the bigger effect comes from routing simple work to free local T0, which this set doesn't isolate.