Gemini 4 Argon API Price Doubles on Dec. 1

Google made Gemini 4 Argon broadly available on October 13 with API pricing of $2 input and $10 output per million tokens that doubles on December 1.

If your store runs AI through an API for product copy, support drafts or feed cleanup, that date is now a budget line. If your buyers shop through Gemini or Google’s AI Mode, Argon is also the model answering them. This is the kind of platform change we track at Ecommerce Paradise because it reaches a store owner from both directions: what you pay to run your back office and what shoppers see before they ever reach your site.

Below: what Google published, how a model that launched to a small group of cyber defenders on September 30 reached everyone in 13 days, what the doubling costs at real volume, and five moves to make before the intro rate ends. My read is that most high-ticket stores will pay pocket change and a few will not. If you are new here, start with my guide to how high-ticket dropshipping works and come back.

Form your LLC with Northwest Registered Agent, the best LLC formation service for ecommerce owners. The AI vendors, payment processors and ad platforms you sign up with all want a legal entity on the account, and Northwest has been doing formations for about 25 years while the cheap formation sites come and go. Start your LLC the right way →

Gemini 4 Argon Goes Broad at $2 and $10 Per Million Tokens

Google published the availability post on October 13, according to Google’s blog. Gemini 4 Argon is now in the Gemini app for Google AI Pro and AI Ultra subscribers. Per 9to5Google, those subscribers can also use it in AI Mode in Search, and Antigravity’s desktop app and command line tool require AI Ultra.

The API numbers come from Google’s pricing page. Standard input costs $2.00 per million tokens through November 30 and $4.00 from December 1. Standard output, which includes thinking tokens, goes from $10.00 to $20.00. Context caching moves from $0.10 to $0.20.

Batch runs at half the standard rate: $1.00 input and $5.00 output now, then $2.00 and $10.00. Priority costs $3.60 input and $18.00 output now, then $7.20 and $36.00. The page lists the model as not available on the Free tier or the Google AI Pro plan, so API use means a paid account.

VentureBeat reported that Google made the model broadly available at 9:08 a.m. PT on October 13, 13 days after the September 30 unveiling. VentureBeat’s comparison puts the intro rate level with Claude Sonnet 5.5 and GPT-6.1 Sol at $2 and $10. Claude Opus 5.5 and Meta Muse Wonder sit at $4 and $20, the same as Argon’s price after the change, and GPT-6 Astra is the most expensive at $10 and $50.

On quality, VentureBeat cites Vals AI, which ranks Argon first on its Vals Index at 68.90%, ahead of Claude Sonnet 5.5 at 67.04% and Claude Opus 5.5 at 66.97%. On the Artificial Analysis Intelligence Index, Argon scores 53, behind Claude Opus 5.5 at 58 and Claude Sonnet 5.5 at 56. Google’s own performance figures are company-reported and have not been independently replicated, per the same article.

Cost per task tells a different story than price per token. Eesel, which sells its own support agent, published Artificial Analysis figures of $1.99 per task for Argon at promo pricing and $3.98 at standard pricing. The comparison points were $0.72 for GPT-6.1 Sol and $1.82 for Claude Opus 5.5. Argon used about 62,000 output tokens per task versus 27,000 for GPT-6 Astra, which is why Argon cost roughly 2.7 times more per task than Sol despite identical $2 and $10 rates.

Google’s post also says Gemini connects to more than 40 apps, including Instacart, OpenTable, Adobe and Shopify, with examples such as ordering groceries and building a business. A new toggle in the main chat window switches between chat and tasks, where Gemini can work on a request while you are away from your computer.

Argon Went From Cyber Defenders to Everyone in 13 Days

Google released Argon on September 30 to trusted cyber defenders in its Fairwind Program, not to paying customers, according to AI Weekly. Paid API access and AI Ultra subscribers were promised later with no date. Google’s self-reported claims at launch included 77.9% on DeepSWE v1.1, 51.3% on AutomationBench for end-to-end business tasks, and output of up to one million tokens, up from 64,000.

The price story shifted in those two weeks. Eesel wrote on October 1 that its API call returned a 404 and that Google had given no end date for the promo, with Artificial Analysis describing it as lasting at least one month. Google’s pricing page now names November 30 as the last day of the intro rate.

There is a counterpoint, and it is not small. AI Weekly reports that Bloomberg heard from some Google employees who say Argon performs worse on real coding tasks than the benchmarks suggest. Google told Bloomberg that characterization is inaccurate. AI Weekly’s editor’s note adds that independent testers flagged a 15% hallucination rate that does not appear in Google’s disclosures.

VentureBeat quotes Sam Witteveen of Red Dragon AI, who tested the model before release: “This is a really, really, really good model. But is it the best model out there? I actually don’t think it is.” He also warned that the expanded token limit is a specialized API capability, not normal behavior in the Gemini app.

Gemini keeps turning up in places that touch a store. I covered Newegg’s Google checkout in Gemini recently. Google’s test of building campaigns from a typed goal, which I wrote up in my natural language campaigns post, sits on the same push.

What Argon’s Price Doubling Means for Your Store’s AI Bill

My read is that the intro rate is a customer-acquisition play, and you should treat $4 and $20 as the real price. Everything below uses that number. The first-order question is volume, so here is hypothetical math, not reported figures.

Say you rewrite 2,000 product pages, each using 1,200 input tokens and 600 output tokens. That is 2.4 million input tokens and 1.2 million output tokens. At promo rates the job costs $16.80. At the standard rate it costs $33.60. Either way it is less than a tank of gas, and the human review of 2,000 descriptions costs far more than the model does.

Batch is the quiet detail. After December 1, batch pricing ($2 input, $10 output) equals today’s promo standard rate, per Google’s page. Anything that does not need an instant answer, like overnight catalog cleanup or ad copy variants, can run at the old price indefinitely.

Agent work is where the bill climbs. Take Artificial Analysis task costs as a stand-in: $1.99 promo and $3.98 standard. At 5,000 agent runs a month that is about $9,950 versus $19,900. Your tasks may cost far less or far more than those benchmark tasks, and I have no data on your mix, but the shape holds: long, multi-step runs burn output tokens, and output is priced at five times input.

Support is the easier case. Eesel’s own estimate, repriced from a test of another model and not run on Argon, is roughly $1.80 to $3 per 1,000 tickets at promo pricing. Doubling that is simple arithmetic: $3.60 to $6. If you use a helpdesk like Gorgias for ticket handling, the model cost is rarely the problem. Wrong answers on a $4,000 order are the problem.

Here are my thresholds. Under about $100 a month in AI API spend, ignore the clock. Between $100 and $1,000, schedule one bake-off before November 30. Above $1,000, run it this week, because a 2x change on that base is a real hire-versus-tool decision.

The other side is demand. Pro and Ultra subscribers can now use Argon in AI Mode, and Google’s own post lists Shopify among the apps Gemini connects to. I already wrote about how AI Mode ads run through your existing campaigns, and the same rule applies here: the model that answers a shopper reads your feed, your images and your prices. Stores with clean data win the recommendation. Stores with blurry images and mismatched prices get skipped, and that is a problem Google’s recent blurry product image decision does not fix.

On which model to use, the counterpoint matters. At the same $2 and $10, VentureBeat’s table shows Claude Sonnet 5.5 and GPT-6.1 Sol priced level with Argon’s intro rate. I run my own AI work through Claude. I also wrote up how to build a leaner high-ticket business with AI. ChatGPT is the other obvious benchmark, so test it too. Do not switch because a launch post says a model ranks first on an index. Switch when it wins on your 50 real tasks.

If this much tooling churn is the reason you have not launched, or your store is live and you have no time to test models, that is what my team handles. The turnkey done-for-you service builds and runs a high-ticket store so you are not choosing between model tiers at midnight.

Want 1-on-1 help deciding where AI belongs in your store and where it does not? Get the coaching details →

Five Moves to Make Before Argon’s Intro Price Ends

Do these in order. The first two take under an hour, and none of them require you to switch models. The point is to know your exposure before the date, so the decision on December 1 is a calm one and not a surprise on a credit card statement.

List every place your business calls an AI model: apps, automations, scripts and plugins. For each one, write the model name and the monthly volume. If you cannot find a number, check the billing page of the tool that calls the model.
Re-price each line at $4 input and $20 output per million tokens in a spreadsheet. Anything under $100 a month gets no further action. Anything over $1,000 gets a bake-off.
Run a 50-item bake-off on one job, such as product descriptions or support draft replies, comparing Argon, your current model and one alternative. Score quality and count output tokens, since output tokens drove the 2.7x cost gap above.
Move anything that can wait onto batch, which Google prices at $1.00 input and $5.00 output now and $2.00 and $10.00 after December 1. Overnight catalog work and email variants for Klaviyo are good candidates.
Fix the data the shopper-side models read: titles, images and prices in your Shopify catalog. Start with the images and make sure Shopify compare-at prices match your feed. A tool like Canva handles a quick image refresh. My post on compare-at prices shown to AI agents explains the mismatch risk.

If you want a second set of eyes on your numbers, book a free call at my discovery page. Bring the spreadsheet from step two.

Frequently Asked Questions

When does Gemini 4 Argon’s intro pricing end?
November 30, 2026, according to Google’s pricing page. From December 1, standard pricing is $4 per million input tokens and $20 per million output tokens.

Can I use Argon on the free Gemini plan?
Google’s blog post does not say whether free users get it. Per 9to5Google, AI Pro and AI Ultra subscribers can use it in the Gemini app and AI Mode, and Google’s pricing page lists it as unavailable on the API Free tier.

Is Argon the best model for ecommerce work?
Unproven. Vals AI ranks it first on its index, while Artificial Analysis ranks Claude Opus 5.5 and Claude Sonnet 5.5 higher. Test it on your own tasks, and see my list of AI tools for dropshipping for jobs that tend to pay off.

Does this change how my store shows up in Google AI Mode?
Not directly. Google did not announce ranking changes, but a stronger model reading your feed raises the cost of messy data. My post on AI Overviews and branded searches covers why branded visibility matters.

Should small stores worry about this price change?
Most should not. A 2,000-product rewrite costs about $34 at the standard rate in my hypothetical above. The exception is long agent runs, where output tokens add up fast.

Where should a new store owner start with AI?
Start with the free high-ticket niches list to pick a category, then add AI where it saves hours, like copy and support drafts. The free mini course covers the foundation.

Will a model switch break my automations?
It can. Prompts tuned for one model often return different formats on another, so test on a copy of the workflow first. Tools like HubSpot that call models in the background deserve the same check.

Want my team to scale the store you already have? See the scaling service →

That is the read for today. I will keep watching the Argon rollout, the November 30 date and any change to AI Mode. Subscribe to the YouTube channel for daily breakdowns. More breaking news coming through the day.

If this was useful, these go deeper:

Trevor Fenner

Trevor Fenner is an ecommerce entrepreneur and the founder of Ecommerce Paradise, a platform focused on helping entrepreneurs build and scale profitable high-ticket ecommerce and dropshipping businesses. With over a decade of hands-on experience, Trevor specializes in high-ticket dropshipping strategy, niche and product selection, supplier recruiting and onboarding, Google & Bing Shopping ads, ecommerce SEO, and systems-driven automation and scaling. Through Ecommerce Paradise, he provides free education via in-depth guides like How to Start High-Ticket Dropshipping, advanced training through the High-Ticket Dropshipping Masterclass, and fully done-for-you turnkey ecommerce services for entrepreneurs who want a faster, more hands-off path to growth. Trevor is known for emphasizing sustainable, real-world ecommerce models over hype-driven tactics, helping store owners build scalable, sellable, and location-independent brands.