Gemini 4 for Marketing: Carbon vs Argon vs Hydrogen, Prices and When to Use Each
Google's Gemini 4 family compared for marketing work: what Carbon, Argon and Hydrogen are for, what each costs per million tokens, and how they appear in MITPO chat.

Google's Gemini 4 line arrived as three models rather than one, and the names do not say which is which. Gemini 4 Carbon, Gemini 4 Argon and Gemini 4 Hydrogen share a one-million-token context window and the same inputs, but they are priced and positioned differently. This guide uses Google's own descriptions and price list (read November 13, 2026) to help a marketer choose, and shows how the three appear in MITPO chat.
We have not run a controlled benchmark of the three on marketing tasks, so this is a positioning and cost guide, not a ranking of output quality.
Key takeaways
- Carbon is Google's most capable reasoning model, for complex software engineering, deep scientific problem solving, high factual accuracy and autonomous multi-step agents.
- Argon is the efficient Gemini 4 frontier reasoning model: faster responses and lower token spend per turn than Carbon on complex coding, agentic, math and science tasks.
- Hydrogen is for everyday coding and agentic workflows, with speed and efficiency, at $1.50 input and $7.50 output per million tokens.
- Shared specs: 1M-token context, up to 65K output tokens (long decodes up to 1M), inputs of text, image, audio, video and PDF.
- Prices per million tokens: Carbon and Argon $2 input and $10 output until November 30, 2026, then $4 and $20. Hydrogen $1.50 and $7.50.
- In MITPO: Carbon and Argon on Growth and above, Hydrogen on Starter and above.
The three models side by side
| Carbon | Argon | Hydrogen | |
|---|---|---|---|
| Google's description | Most capable reasoning model | Efficient frontier reasoning model | Speed and efficiency for everyday coding and agentic work |
| Model code | gemini-4-carbon | gemini-4-argon | gemini-4-hydrogen |
| Context window | 1M tokens | 1M tokens | 1M tokens |
| Max output | 65K, long decodes to 1M | 65K, long decodes to 1M | 65K, long decodes to 1M |
| Inputs | Text, image, audio, video, PDF | Text, image, audio, video, PDF | Text, image, audio, video, PDF |
| Input price per 1M tokens | $2.00 to Nov 30, then $4.00 | $2.00 to Nov 30, then $4.00 | $1.50 |
| Output price per 1M tokens | $10.00 to Nov 30, then $20.00 | $10.00 to Nov 30, then $20.00 | $7.50 |
| Cached input per 1M | $0.10, then $0.20 | $0.10, then $0.20 | $0.075 |
| Free tier | No | No | Yes, with limits |
Two billing details matter for marketing teams. Thinking tokens are billed as output, so a reasoning-heavy answer costs more than its visible length suggests. And Google only supports implicit caching on these models, so repeated long briefs get a discount automatically when the prefix repeats, but you cannot set it up yourself.
Which one for which marketing job
- Carbon: the hard, high-stakes pass. Long-document analysis, a full competitor teardown from several PDFs and videos, or a multi-step plan where a wrong fact is expensive.
- Argon: the same family, tuned for speed and spend. Google says it gives Gemini 4 intelligence with faster replies and lower token spend per turn than Carbon. Choose it for the long middle of a working session.
- Hydrogen: volume. Drafts, variants, tagging a content library, summarising calls and video, and agent runs that fire often.
A practical split: draft and iterate on Hydrogen, switch to Argon for the careful rewrite, and keep Carbon for the one pass where accuracy matters most.
Multimodal input is the real marketing feature
All three accept video and audio as well as text, images and PDFs. That is useful for work that used to need a person to watch something:
- Upload a competitor's ad reel and ask for the hook, the offer and the call to action in each cut.
- Hand over a webinar recording and ask for five LinkedIn posts with timestamps.
- Give a product demo video and a brand guide, and ask what in the demo breaks the guide.
The one-million-token window means a long brief plus several reference documents fits in a single request.
Three prompt templates
1. Competitor reel breakdown
Watch the attached video. For each cut give: time, what is shown, the hook or claim,
and the call to action. Then list the three patterns that repeat. Quote only what is
said or shown on screen; mark anything you are inferring.2. Long-document brief
Read the attached documents. Produce a one-page brief with: audience, the single
promise, three proof points with the page they come from, and what is missing.3. Variant batch
Write 12 headline variants for [offer]: four that lead with the benefit, four with the
objection, four with a number. Max 8 words each. No exclamation marks.Limits to plan around
- Prices change on December 1. Carbon and Argon double on that date, and any budget built on the November rate is off by half.
- No quality ranking here. The positioning above is Google's; run your own brief across two models before you standardise.
- Facts still need checking. A model with high factual accuracy as a goal can still be wrong; check names, numbers and quotes against the source.
- Not available on Google's free tier. Carbon and Argon need paid access; Hydrogen has a limited free tier.
Gemini 4 in MITPO
In MITPO chat, open the model picker and choose Gemini 4 Carbon or Gemini 4 Argon (Growth and above) or Gemini 4 Hydrogen (Starter and above). Chat draws from your plan's usage allowance rather than a per-message price, and the picker shows which plan each model needs. See the pricing page for plan limits and our guide to reasoning models in marketing for how to choose between model families.
FAQ
What is the difference between Gemini 4 Carbon and Argon?
Google calls Carbon its most capable reasoning model and Argon its efficient Gemini 4 frontier reasoning model, with faster responses and lower token spend per turn than Carbon. They cost the same per token.
How much does Gemini 4 cost?
Carbon and Argon are $2 input and $10 output per million tokens until November 30, 2026, then $4 and $20. Hydrogen is $1.50 and $7.50.
What is Gemini 4 Hydrogen for?
Everyday coding and agentic workflows where speed and cost matter, with the same one-million-token context window as the larger models.
Can Gemini 4 read video and audio?
Yes. All three accept text, image, audio, video and PDF input.
Which Gemini 4 model should I use in MITPO?
Start on Hydrogen for drafts and volume, move to Argon for careful rewrites, and use Carbon for the hardest analysis.
Put this to work in MITPO
Open the demo with no account, or create a free account and start with your own brand.
MITPO Editorial & Research Team
Operator-verified playbooks benchmarked on active 2026 marketing workflows.
Further Reading
Related Articles

The 2026 AI Price War: What Opus 4.8, Gemini 3.5 Flash and Microsoft MAI Mean for Your Marketing Budget
Claude Opus 4.8, Gemini 3.5 Flash, and Microsoft's MAI models just kicked off an AI price war. Here's the right model stack for each marketing job - and what it actually costs a lean team.

Claude Fable 5: What Frontier AI Models Actually Mean for Marketing Teams in 2026
Claude Fable 5 is here. What Anthropic's newest frontier model really changes for marketing teams — and how to evaluate any new AI model in 30 minutes.

The Future of AI in Digital Marketing: 2026 Trends You Can't Ignore
A 2026 operator brief on AI search, agentic workflows, first-party data, creative automation, and the systems marketers need to stay competitive.