You're Probably Using the Wrong AI Model
The one thing you need to know in AI today | AI-Ready CMO
We don’t usually cover model launches. Most of the time, a new release means marginally better benchmarks, a few breathless LinkedIn posts (we love them — no, not really), and absolutely nothing changes about how you do your job on Monday.
However, just in the last 2-3 weeks, Anthropic, Google, and several Chinese labs have dropped major releases almost simultaneously. Rather than walking you through each one’s benchmark scores (you can safely ignore those), we want to focus on what actually matters: these models aren’t competing with each other anymore. They’re competing for different jobs.
The era of “which AI is best?” is over. The question is now “best for what?”
Let’s start with the one we use every day.
Anthropic’s new Claude Sonnet 4.6 now has a 1-million-token context window, and in practice, this changes what the tool actually is. It’s not a text generator you prompt and hope for the best. It can hold an entire campaign brief, your brand guidelines, two years of performance data, and a competitive analysis in a single conversation — and actually reason across all of it.
If you run global marketing or CX operations and you want the lowest-friction path to agentic workflows — AI that actually does things across multiple steps — this is your current best option. It’s also a great copywriter.
Google finally showed up.
Gemini 3.1 Pro is the next flagship from Google Workspace. A million-token input, 64K output, and genuinely strong reasoning and coding. What matters for us: it handles text, image, audio, and video natively. No new subscriptions, no migration headaches.
And in our experience, no model comes even close when working across languages. Gemini 3.1 Pro can write you English, French, Japanese, Armenian, or Kazakh copy with equal ease.
Now, the part most Western marketers are ignoring — and shouldn’t be.
The real structural shift is happening in Chinese AI labs. Models like GLM-5, Qwen 3.5, Kimi-K2.5, and MiniMax-M2.1 might not make it to your LinkedIn feed, but they’re changing the economics of AI-powered automation. Not to mention the latter two are just genuinely pleasant to talk to — try them out!
These models are absurdly cheap to use.
We’re talking fractions of a cent per task — the kind of pricing that makes it viable to embed AI into simple, repetitive workflows where Claude or Gemini would be overkill (and overpriced). Think: automated lead scoring, bulk content tagging, first-pass translations, or routing customer inquiries before a human ever sees them. The kind of work that doesn’t need frontier-level reasoning, just reliable execution at scale.
Six months ago, “cheap” meant “noticeably worse.” That gap is closing fast. These models are now maybe a few months behind their Western counterparts in quality — and for the straightforward automation tasks most marketing teams actually need, that gap is already irrelevant.
If you’re paying premium API prices to have GPT-5.2 or Claude Opus do work that doesn’t require their full capabilities, you’re overspending. The smart play is matching the complexity of the task to the cost of the model.
So what do you actually do with this?
Honestly, most of you don’t need to do anything dramatic right now. You don’t need new subscriptions or to force your engineering colleagues to spin up anything by Friday.
But you do need to stop treating “AI” as a single thing. You don’t want to use a $20-per-task model for a $0.002-per-task job — or vice versa.
Producing multimodal or multilingual content? Gemini 3.1 Pro.
Need agentic workflows with minimal infrastructure or just a really smart generalist? Claude Sonnet 4.6.
Running bulk automation that needs to be fast, reliable, and cheap? The Chinese models are your answer.
The market has been segmented, and the gap between getting this right and getting it wrong is costly, either in time or in quality. Time to have a model strategy.
— Torsten and Peter
3 AI Marketing Tools To Try Today
Doodle
Find a meeting time that works for everyone without the email ping-pong—Doodle polls your group and locks in availability automatically.
Comp AI
Get SOC 2 ready in 24 hours instead of 24 weeks — Trycomp auto-collects evidence across AWS, GitHub, and Slack without hiring a security team or paying Vanta prices.
Foxy AI
Generate photorealistic images and videos of yourself without a camera or crew—Foxy AI creates hyper-realistic content from uploaded photos so creators scale output without scheduling shoots.
Know Where You Actually Stand With AI.
Your team is experimenting with AI tools. Some are using ChatGPT for drafts. Others ignore it completely. You’re not sure if you’re falling behind or ahead—and you definitely don’t know what to prioritize next.
The AI Marketing Readiness Assessment gives you a clear diagnosis in 5-10 minutes. Answer 25 questions calibrated for marketing teams.
Get instant scoring across strategy, skills, infrastructure, and adoption.
See exactly where you’re strong, where you’re exposed, and what to fix first—whether you’re a solo marketer, team leader, or CEO building AI capability.
No generic advice. No consultant upsell. Just your current state and your next steps.
Available for AI-Ready CMO paid subscribers.