· 5 min read

Gemini 4 Argon Doubles in Price on December 1. Model Your Margins at $4 and $20, Not at the Launch Rate.

AI Tools

Gemini 4 Argon became broadly available on October 14, and the number most coverage led with was $2 per million input tokens and $10 per million output tokens. Google's own pricing page shows the second row that matters more: from December 1, 2026, those become $4 and $20. If you are choosing a model for a product that will still be running in January, the launch price is the wrong number to build a margin on.

The whole price table doubles

The paid-tier Standard rates, per million tokens, run $2.00 input and $10.00 output through November 30, then $4.00 and $20.00. Output includes thinking tokens, which is where agentic workloads quietly spend most of their money. Context caching goes from $0.10 to $0.20, and the page says caching is implicit only, so there is no explicit cache control to tune.

The other tiers move the same way. Batch and Flex are $1.00 input and $5.00 output now, then $2.00 and $10.00. Priority is $3.60 and $18.00, then $7.20 and $36.00. There is no free tier for this model. So after December 1, the Batch and Flex rates equal today's Standard promo rate, which is the one piece of good news in the table.

What that does to a real bill

Take a hypothetical product that moves 10 million input tokens and 2 million output tokens a day, about 300 million and 60 million a month, with nothing cached. At the promo rate, that is $600 for input and $600 for output, $1,200 a month. From December 1 on Standard, it is $2,400. Add caching and the shape does not change, because every line doubles together.

If you charge customers a flat $29 a month and have 100 of them, you can see the problem: the promo bill is about 41% of that $2,900 revenue and the real bill is about 83%. I picked round numbers on purpose. Plug in your own and the point survives: a margin that only works at $2 and $10 is not a margin.

The window detail that changes long prompts

The New Stack reports that Argon has a single 1 million token limit shared by input and output, so a long prompt leaves less room for the answer. The September 30 announcement had described the limit as a 1 million token output cap, up from 64,000 in earlier Gemini models, so check what your own requests actually get before assuming a long document plus a long answer fits. For summarization over big inputs, this rarely bites. For agents that stuff history into every call and also want long outputs, it can.

Availability, since it keeps changing the answer

API access is through Google AI Studio and the Interactions API as gemini-4-argon, on the paid tier. In Google's own Antigravity tool, Argon is available to AI Ultra subscribers and eligible Gemini Enterprise plans, not AI Pro. If you were planning to prototype on a Pro subscription and ship on the API, the two will not behave identically.

What I'd actually do

Price the product on the December rate today. Put $4 and $20 in your cost spreadsheet, and treat the difference between that and what you actually pay in November as profit you have not earned yet, not as margin to spend.

Move anything that is not user-facing to Batch or Flex. After December 1 those tiers will cost what Standard costs this month. Read the latency terms on the pricing page first, because a cheaper tier that makes a nightly job miss its window is not cheaper.

Run your ten most expensive real prompts on Argon and on whatever you use now, before the change, and log cost per completed task, not cost per token. Per-task cost is the number that survives a price change, since a model that thinks longer can lose its price advantage even at a lower rate.

Where this could be wrong

Google can change this page at any time. Promotional windows get extended, and competitors may respond with their own cuts, which would make building on the December rate look overcautious. Pricing the worst case costs you nothing if the best case arrives. Pricing the best case costs you a margin if it doesn't.

Author

Lukas

Sources

Gemini Developer API pricing, Google AI for Developers Gemini 4 Argon debuts with a unified 1M input/output token limit, The New Stack Top Tech News Today, October 14, 2026, Tech Startups

More AI articles →

Newsletter

The week, in one email.

The weekly summary of what happened and what it means for people who build alone. No spam, unsubscribe any time.

Form not showing? Subscribe on this page instead.

Related Posts

AWS Shipped a Free Local Emulator That AI Agents Keep Up to Date. Telemetry Is On by Default and IAM Is Not Checked. Here Is How I'd Use It.

AWS Emulator runs AWS APIs on your laptop with no account and no fee, and it hit about 785 points on Hacker News. It also skips IAM policy checks, keeps state in memory, and sends usage data to AWS unless you switch it off. Good for fast tests, wrong for security tests.

Apple's Claris Is Testing Harbor, a Private AI Server That Pools Memory Across Macs Over Thunderbolt 5. Here Is What It Would Replace in My Setup.

Claris, an Apple company, has a private preview page for Harbor: run open-weight models on Macs you own, and pool unified memory across two or more Thunderbolt 5 Macs. No price, no named models, no API spec. Here is what is confirmed, what is not, and the math to run before you buy hardware for it.

Google's Gemini Decisions API Scores Up to 1,000 Questions per Request at $0.10 per Million Input Tokens. It Changes Yesterday's Self-Host Math.

Google launched a hosted decision model on November 11 that returns probabilities instead of text, bills only input tokens, and lets one request answer up to 1,000 questions. Against a rented GPU, the self-host break-even jumps from about 1,500 calls a day to tens of thousands. Test the accuracy before you switch.