The Tin Can Is Dead: Why Gemini 3 Finally Matters for Marketers

Everyone is going to obsess over the benchmark scores today.

Peter Benei and Torsten Sandor, 19 November 2025

Everyone is going to obsess over the benchmark scores today. They will talk about ELO ratings, math proofs, and Python coding capabilities.

You can safely ignore most of that.

After living with Google’s newly released Gemini 3 for the last 8 hours, the headline isn’t about physics or calculus. It’s that Google finally killed the “tin can” effect.

For the first time, Google’s flagship model doesn’t sound like a press release tailored by a committee of risk-averse lawyers. Or a tin can dressed up in a cheap suit playing corporate spokesperson. The jump from Gemini 2.5 to Gemini 3 is violent, especially if you use style samples it can follow.

We historically avoided Gemini for long-form copy because it had a distinct, metallic aftertaste—safe, boring, and utterly uninteresting AI slop. That is gone. Gemini 3 feels natural, human, and surprisingly distinct.

While I’m not ready to say it fully dethrones Claude Sonnet 4.5 as the king of stylistic nuance (Claude still has a bit more soul, or “smell” as they say), Gemini has firmly moved from “unusable for creative work” to “essential daily driver.”

Where it truly leaves everything else in the dust is multilingual work. If you manage global campaigns, this is your new engine. We threw obscure languages and complex colloquialisms at it, and it didn’t just translate; it transcreated. It writes compelling, native-sounding copy—the kind of work that used to require expensive local agencies or hope that your Bulgarian freelancer understood your brand voice. It is undefeated in this category.

And it is free, sort of.

Sure, there is a $20 per month Pro plan but through AI Studio, Gemini 3 is essentially free with limits so generous you will likely never hit them doing standard marketing work. While competitors are metering your usage and gating their best reasoning models behind higher paywalls, Google is flooding the zone. You can currently run your entire content operation on the state-of-the-art model for zero dollars.

One more thing: the widely rumored image model—referred to in leaks as either Nano Banana Pro or Nano Banana 2—was noticeably absent from yesterday’s drop. I don’t think the week is over at Google. If the leap in visual generation matches the leap we just saw in text, your creative workflow is about to get another shock. For now, update your prompts, switch your default model in the settings, and enjoy the feeling of a Google model that finally understands nuance.

3 AI Marketing Tools To Try Today

RankPrompt

→ Try it here

Discover what prompts actually work for your use case—RankPrompt tests prompt variations, ranks results, and shows you the highest-performing combinations.

Quillbot

→ Try it here

Rewrite and refine content without starting from scratch—Quillbot paraphrases, summarizes, and improves grammar instantly for cleaner messaging.

Apollo

→ Try it here

Find and engage B2B prospects at scale—Apollo combines contact data, email sequences, and engagement tracking to automate your entire outbound pipeline.

Help us send you better content (30 seconds)