You only need the frontier model for one single edit

You only need the frontier model for one single edit — Stencil

Monkey see, monkey do! 🍌 97%of frontier performance 41%cheaper in $$ 1.9×faster completion ~3×less likely to cheat

Cost vs pass rate, 7 arms · SWE-Bench Pro · a bare model = oneshot · $/task includes the frontier model's opening turns · † executes with Flash 3.5 · ‡ executes with 5.6 Luna /plan makes perfect sense. It really shouldn't! You've heard this pitch; you may have even shipped it. The expensive model is clearly the better architect, but it feels like a waste to use it for the entire pipeline. Why not let it do the "hard part"? Read the code, think deeply, write a precise plan. Then a model a tenth the price executes the plan. Senior architect, junior engineer. Sounds great, right? Find the red dot above, labeled Opus 4.8 + /plan†. Opus plans read-only, Gemini Flash implements: lands at $3.18 per task, 12.7 minutes, 84.6% pass . Opus doing the entire task by itself, no handoff, no junior: $2.78, 10.1 minutes, 84.6% . The "cost-saving" measure costs 14% more than not saving. Huh? reads it all · $$$reads it all again · $$ code · ~100K tokens

opus · plans

plan.md · a 2K postcard

flash · implements

the patch · ~2K tokens

The mistake is upstream of the architecture diagram. People price agents the way they price people: senior time is expensive, so minimize senior involvement. But the expensive part of an agent's day is not the fixing, building, or even the thinking. Opus fixing things does not cost money. Opus reading things costs money. Take a look at this admittedly anecdotal distribution of where my tokens went; fully automated agents look no different. 1.81B tokens across ~2M tool calls · "doing the task" (every edit and write) is 9%; reading is what scales the bill, and both models pay full price for it Nine percent of the tokens are edits. The rest is reading, and this split is not a quirk of one harness, or something you can "fix". Trust me, I've tried with snapcompact. Any agent, any model, any scaffold: the bill is essentially O(reads). Now walk through every reason you'd reach for /plan, with that in mind:

"I want the deep understanding of the big model." The understanding lives in 100K+ tokens of grounded context: files read, dead ends eliminated, hypotheses tested. The plan document is a 2K-token postcard from that context. The executor gets the postcard, not the understanding, and has to rebuild the rest at its own expense.

"The task is very complicated." Then you don't want the main agent executing at all; a single read-only planning turn isn't the answer. Let it explore, then dispatch sub-agents to do the work. A game of telephone doesn't help you here.

"I'm cost constrained." Reading is the cost. /plan makes the frontier model read everything at frontier prices, then makes the cheap model read it again. You didn't move the expensive part; you duplicated it.

Here's what that looks like in practice, with a diagram I've spent way too much time on: OPUS 4.8 + /PLAN† · $3.18 bashopus globopus readopus bashopus bashopus readopus bashopus readopus readopus readopus grepopus readopus readopus grepopus grepopus readopus readopus readopus ¶¶proseopus readopus

recon: sessions ▸ signing

writeopus resolveopus ▸ flash

plan

readflash readflash editflash readflash editflash

edits

readflash basherrorflash grepflash editflash grepflash readflash editflash bashflash bashflash ¶¶proseflash

verify ▸ debug ▸ pass

Σ 1.34M

OPUS 4.8 · SAME TASK · $2.78 bashopus readopus readopus bashopus bashopus readopus grepopus readopus readopus bashopus readopus grepopus

recon: sessions ▸ signing

editopus readopus editopus

fix

readopus editopus bashopus grepopus readopus grepopus readopus editopus

tests ▸ debug warnings

bashopus bashopus bashopus readopus ¶¶proseopus

verify ▸ close

Σ 1.10M

OPUS 4.8 + /PREWALK† · $1.46 bashopus readopus nudge readopus bashopus bashopus readopus readopus bashopus readopus ¶¶proseopus todoopus

recon ▸ plan

editopus ▸ flash

fix

readflash editflash todoerrorflash todoflash

tests

basherrorflash grepflash grepflash readflash readflash editflash bashflash

debug warnings

grepflash readflash grepflash grepflash bashflash todoflash todoflash ¶¶proseflash

checks ▸ close

Σ 1.13M opus flash read write exec todo ¶ prose error harness event Σ tokens

django-13279 test run † executes with FlashLook at the top ribbon. Opus reads base.py, signing.py, the test file (twenty cards of gray), then writes its plan and leaves. And what's the first thing Flash does with that beautiful document? It re-reads base.py and the test file, because a plan is not a file and you cannot edit prose. The gray reads just keep stacking, first at Opus prices, then again at Flash prices. There is no version of this where a second reader is the cost optimization. Hand off a trajectory, not a fairytale A plan document is a literal postcard, describing a journey to a model that never took it. What actually could transfer something of...

You only need the frontier model for one single edit

Related Articles

(no title)

Scientists reverse brain aging, with a nasal spray

AI has torched the market for junior programmers

Al Vigier: Canada's AI strategy shouldn't include secret Palantir bills

The Anatomy of an AI-Native Org