cost & savings

The arithmetic, with the awkward parts left in.

Every number on this page is either a published list price or a measurement from one real week on the Triss codebase. Where Triss does not pay off, that is on the page too.

estimate your own billtoken profile: 28K in / 4.7K out per request

Agent requests per day40

Share delegated to Triss

Measured week ran at roughly two thirds — bulk reads, reviews, fetches, commit messages.

Primary model

Worker

DeepSeek off-peak: 01:00–04:00 and 06:00–10:00 UTC peak, rest off-peak 50% off — calculator uses off-peak.

Cache hit rate77%

average 90–95% now, week 18.5% overall (28.9% on pro)

Agent alone$185

Agent + Triss$75.55

saves $110 / month · 59% less

annual $1,318 saved

One measured week — not a projection

209 calls on the Triss codebase itself. Flash for bulk reads, pro for reviews.

metric

pro

flash

total

Requests

143

209

Input tokens

3.74M

2.10M

5.84M

↳ cache read

1.08M

256

1.08M

↳ cache write

—

Output tokens

833K

156K

990K

Cost (USD)

$1.88

$0.34

$2.22

Cache read 1.08M — 18.5% of input tokens were cache hits (mostly on pro). Without cache the week would have cost ≈$2.69 instead of $2.22. Cache is the main lever after delegation.

DeepSeek pricing — off-peak USD per 1M (peak 2×) — 2026-08

Source: DeepSeek's official pricing notice and price card; effective 2026-08-16 16:00 UTC. CLI source: src/usage.js DEEPSEEK_PRICING — override via TRISS_PRICE_<MODEL>. Peak: 01:00–04:00/06:00–10:00 UTC; rest off-peak.

metric

flash — $0.22 / $0.007 / $0.66

pro — $0.66 / $0.022 / $1.98

input uncached / 1M

$0.22

$0.66

cache read / 1M

$0.007

$0.022

output / 1M

$0.66

$1.98

Calculator is honest: workerCost = IN*(1-hit)*inputPrice + IN*hit*cachePrice + OUT*outputPrice — previously cache was ignored and averaged as $0.0052/$0.0131. Cache write for DeepSeek = $0.

Worker avg $0.0106/call measuredDeepSeek flash off-peak $0.22/$0.007/$0.66 per 1M (peak $0.44/$0.014/$1.32)

Tiny lookups

A two-file read roughly breaks even: the round trip plus the summary costs about what doing it directly would.

Judgement calls

Architecture decisions, ambiguous requirements — being wrong is expensive. Cheap models are for volume, not verdicts.