
Opus for planning. Sonnet for writing. M2.7 for grunt work.
The marketer who doesn't route is the marketer paying 8x.
Costs matter. Last month, I cut my model bill by 64%. The change was structural, not tactical. I stopped letting Claude or Codex pick.
The cost differential is bigger than you think. Opus 4.7 runs $25 per million output tokens. MiniMax M2.7 is $0.03 (token plan). That's an 800x spread, top to bottom. In between, in cost order: Sonnet 4.6 ($15), Gemini 3.1 Pro ($12), Haiku 4.5 ($5), Grok 4.3 ($2.50 with a 1M context window).
Absolute prices keep dropping. The 800x structure doesn't.
Most of us use one provider. Whatever's the default. Usually the expensive one.
I cut my own bill 64% last month by smarter routing. Same workflows. Same outputs. The difference was deciding to stop defaulting.

Giphy

Top to bottom by cost (output $/M):
Opus 4.7 ($25) — reasoning. Plan reviews. Architecture decisions. Strategic synthesis across a lot of context. The judgment calls. Don't use it for regular work.
Sonnet 4.6 ($15) — drafting, editing, customer-facing writing. Anything that needs voice and nuance but not deep reasoning. Your workhorse for content. Sonnet wrote most of this newsletter.
Gemini 3.1 Pro ($12) — Research. Native multimodal makes it the call for image work: asset audits, brand consistency checks, screenshot review. The research side is Google-native, so it pulls signal other can't.
Haiku 4.5 ($5) — classification, scoring, formatting. The mechanical parts of a workflow. Tagging emails. Categorizing leads. Running a quality rubric. Anywhere you're checking against a list and security matters.
Grok 4.3 ($2.50, 1M context) — SEO research and social research. Competitor teardowns. Thread-mining. SERP analysis. The big context window means you feed it everything and ask it to surface patterns.
MiniMax M2.7 ($0.03, year token plan) — bulk. Where Haiku felt cheap, MiniMax feels free. Mass classification, signal scoring, tagging at volume. Anywhere you're running the same job across thousands of records. You can run it locally, too on your own machine.
What this looks like in practice
I run a content pipeline that has six stages: ideation → outline → draft → quality check → voice polish → publication formatting.
Three of those stages run on Haiku. Two on Sonnet. One on Opus.
Ideation (Haiku) — generating candidate topics by scanning signals. Mechanical, list-driven, no judgment yet.
Outline (Opus) — the only Opus step in the chain. Structure decisions, competitor analysis, SEO judgment. This is where reasoning actually matters.
Draft (Sonnet) — turning the outline into prose. Voice work, not reasoning.
Quality check (Haiku) — scoring against a 6-dimension rubric. Comparison, not creation.
Voice polish (Sonnet) — applying MA voice. Pure writing.
Publication formatting (Sonnet) — channel-specific formatting (newsletter, blog, social). Pattern application.
One Opus step. The other five are 5-60x cheaper. The output is actually better. I did extensive testing for over a month.

NEW! Marketing AI: The Future of Digital Marketing
Turn AI Into Your Most Powerful Marketing Teammate
DAY 1: Wed Jul 29, 2026 | 9:00 AM - 12:00 PM | Live Online
DAY 2: Fri Jul 31, 2026 | 9:00 AM - 12:00 PM | Live Online
Put AI to Work Across Your Marketing Strategy
In this hands-on course, you will learn how to confidently use tools like ChatGPT, Midjourney, and Descript to plan, create, and scale content across copy, images, and video.
You will leave with practical frameworks and real experience generating marketing assets you can apply immediately.
Led by Alec Newcomb, this course is built for marketers, leaders, and business owners who want to improve efficiency, drive growth, and stay ahead with AI.
The two pitfalls
Opus for everything. You pay 5x. You don't get 5x. Opus is overkill for most marketing work, the bill isn't.
Haiku for synthesis. The test is embarrassingly simple.
If the task is "check this against that list" or "format this in this shape" — Haiku.
If it's "decide what matters here" — Opus.
If it's "say this well" — Sonnet..
The 80/20
Roughly 70-80% of marketing AI work is mechanical. Lead scoring. Email categorization. Draft tagging. Quality checks. Ad Reports. SEO audits. Content classification. 15-25% is writing and applied research. Drafts. Replies. Polish. Competitor teardowns. Image review. Sonnet, Grok, Gemini.
5% is real strategic reasoning. The plan reviews. The "should we do this?" calls. The synthesis across 40 documents. That's where Opus or ChatGPT 5.5 earns its bill.

Model routing isn't a clever prompt or a tool feature. It's an operational habit. The same way you don't run paid traffic without a Google Analytics UTM’s, you don't run an AI workflow without a routing strategy.
Start by listing every task you ask AI to do this week. Tag each one: reason, write, or check. Then look at where your money actually goes. The mismatches are the savings.

This week’s FREE Resource: The Marketer’s Guide to AI Model Routing
Your guide to choosing the right AI model for writing, research, reasoning, and bulk workflows without overspending.
Cut Your AI Costs Without Lowering Output.

What's your highest-spend task right now?
— Alec




