Gemini 4 Argon pricing is introductory. Google charges $2.00 per million input tokens and $10.00 per million output tokens through 30 November 2026, then $4.00 and $20.00 from 1 December. If your team calls the API, that's a bill that doubles in seven weeks. If you run GitHub Copilot Business or Enterprise, the model may already be switched on for your developers, and GitHub's docs mark its Copilot price as promotional too. Anyone who budgets off the launch-week numbers will be wrong by December.
Google's changelog lists gemini-4-argon as generally available from 13 October 2026 and calls it "our most capable model built for complex coding, agentic, math, and science tasks." For other launches, see our AI news hub.
What Gemini 4 Argon costs, and when the price changes
Google's Gemini API pricing page lists the rates per million tokens. Output includes thinking tokens.
| Tier | Input to 30 Nov | Input from 1 Dec | Output to 30 Nov | Output from 1 Dec | Cached input to 30 Nov | Cached input from 1 Dec |
|---|---|---|---|---|---|---|
| Standard | $2.00 | $4.00 | $10.00 | $20.00 | $0.10 | $0.20 |
| Batch and Flex | $1.00 | $2.00 | $5.00 | $10.00 | $0.05 | $0.10 |
| Priority | $3.60 | $7.20 | $18.00 | $36.00 | $0.18 | $0.36 |
Every row doubles, and the multiplier is the same on input, output and cache. Moving to Batch or Flex halves the rate, but it doesn't change the 1 December date. Search grounding is priced separately: 5,000 free search requests a month, shared across Gemini 3 and newer models, then $14 per 1,000 requests, according to the same page.
Google's launch post for Argon repeats the date in a footnote: after the introductory period expires on 30 November 2026, the $4 per 1M input and $20 per 1M output price applies. The post also says cached input tokens are 95% off.
Access is narrower than the headline price suggests. Google's pricing page says Argon isn't available on the Free tier or on the Google AI Pro plan. You need a paid API setup.
Caching deserves a flag. The pricing page says "Implicit caching only (explicit caching is not supported)." You can't pin a large context and pay the cached rate on purpose. Whatever cache discount shows up depends on Google's automatic behaviour, so don't build a cost model that assumes a specific hit rate.
What the doubling does to a monthly bill
Here's a hypothetical, not usage data. Say a coding agent burns 100 million input tokens and 20 million output tokens a month on Standard. The volumes are round numbers picked to keep the arithmetic readable.
Through 30 November: 100 x $2.00 = $200 for input, 20 x $10.00 = $200 for output. Total $400.
From 1 December: 100 x $4.00 = $400 for input, 20 x $20.00 = $400 for output. Total $800.
Same workload, $400 more a month, from nothing but the calendar. Scale the volume tenfold and the gap is $4,000 a month. The point is that a trial run in October says very little about what the same workload costs in January. Your own usage dashboard has the real token counts, and those are the numbers to put into the table above.
How Gemini 4 Argon is billed in GitHub Copilot
GitHub's changelog for the launch says Argon is available to Copilot Pro+, Max, Business and Enterprise users, and that the rollout will be gradual. It also says the model "is billed at provider list pricing under usage-based billing." The rules behind that are in Copilot's usage-based billing rules, and we won't repeat them here.
The enablement part matters more for admins. Business and Enterprise admins manage access through the model policy in Copilot settings. Under Copilot's default enablement policy, new models are on unless an administrator has turned off the global default or disabled the model. So if nobody touched those settings, Argon becomes available to your developers as it reaches your org, and any use of it is billed at list price.
GitHub's models and pricing documentation lists Argon at $2.00 per million input tokens, $0.10 for cached input and $10.00 for output, with 1 AI credit equal to $0.01. A footnote marks that pricing as promotional "until December 1, 2026."
What neither GitHub page tells you is the price after that date. The changelog doesn't mention the Google price change at all, and the docs give no post-promotional figure. We're not going to guess one. Treat it as an open question and ask your GitHub contact, or check the models and pricing page again before 1 December.
What this means for your AI tooling budget
What follows is our judgment, built on the dates above.
For API use, trial Argon at the introductory price if you want to, but model the production workload at $4/$20 from the start. A pilot that only works at $2/$10 isn't a pilot you can afford to keep. Put 1 December on the finance calendar now.
For Copilot Business and Enterprise, someone with admin rights should make a deliberate choice before 1 December: keep Argon on, set a budget around it, or disable it in the model policy. Leaving it to the default is also a choice, just one nobody made on purpose. The same Copilot admin call for GPT-6.1 Sol is open too, so settle both in one pass. Gemini 3.8 Flash's earlier price doubling followed the same pattern with a January date, which makes Argon the second Google launch with the same structure.
To decide whether the spend earns its keep, a plan a VP can run to measure AI coding tools ROI is the place to start.
Frequently asked questions
When does Gemini 4 Argon's price go up?
On 1 December 2026. Through 30 November, Standard costs $2.00 per million input tokens and $10.00 per million output tokens. From 1 December it's $4.00 and $20.00, according to Google's pricing page and launch post.
Is Gemini 4 Argon free or on Google AI Pro?
No. Google's pricing page says Argon isn't available on the Free tier or on the Google AI Pro plan. You need paid API access.
Is Argon turned on automatically in GitHub Copilot?
Yes, by default. GitHub's changelog says new models are on unless an administrator has turned off the global default or disabled the model, and the rollout is gradual. Argon is available on Pro+, Max, Business and Enterprise plans.
Does explicit context caching work with Argon?
No. Google's pricing page says only implicit caching is supported and explicit caching isn't. Cached input is priced at $0.10 per million tokens through 30 November and $0.20 from 1 December on Standard.