Gemini 4 Carbon starts with trusted defenders and time-limited introductory pricing

Google has announced Gemini 4 Carbon, with access beginning for a subset of cyber defenders rather than a general developer or consumer release. The October 28 announcement says the company will expand access gradually to additional trusted partners through its Fairwind Program, while continuing work on guardrails.

The practical distinction is between a model announcement and an available integration. A developer planning a switch cannot treat this release as confirmation that Carbon is enabled on their account, that a public model identifier is ready to use or that an existing Argon integration automatically supports it. The announcement does not provide those activation details. Its stated next audience is selected partners; broader product access remains ahead.

Google publishes an introductory launch price of $2 per million input tokens and $10 per million output tokens. A footnote to those rates says that after the introductory period expires, the price will be $4 per million input tokens and $20 per million output tokens. The end date of the introductory period is not specified in the announcement.

Announced price period	Input, per 1 million tokens	Output, per 1 million tokens
Introductory period	$2	$10
After the introductory period	$4	$20

These are separately priced input and output quantities, not a flat charge for a million‑token conversation. A budget needs both volumes and the applicable period. The announcement also advertises cached input at 95% off the input‑token price. That does not establish that every input will qualify for caching or that a reader has received the discount on an actual bill. No account or billing test was performed for this report.

The [release](https://blog.google/innovation‑and‑ai/models‑and‑research/gemini‑models/introducing‑gemini‑4‑carbon/) describes Carbon as coming to developers, businesses and consumers through Google’s product surfaces. It does not give a general availability date, a consumer plan entitlement or a complete list of surfaces at launch. Google also says it is participating in the US government’s voluntary process for pre‑release model testing. That statement is not an approval or an independent guarantee of security. For an organization evaluating Carbon, model access, permitted use and production deployment are separate checks. A team would need to establish its eligibility and the supported endpoint before planning a migration, then check the terms and pricing that apply to its actual use. Neither an announced future price nor a selected partner’s access establishes availability for every customer.

Google positions Carbon for sustained engineering workflows and cybersecurity defense. The [launch post](https://blog.google/innovation‑and‑ai/models‑and‑research/gemini‑models/introducing‑gemini‑4‑carbon/) also contains benchmark and internal evaluation claims. This report does not repeat their scores or infer real‑world security effectiveness: the underlying benchmark methods, safety report and customer deployment results were not independently reviewed here. The useful confirmed boundaries in the announcement are narrower: a selected defender‑first rollout, a planned expansion to other audiences, and introductory token prices with a separately stated later rate. Those details let readers distinguish what Google has announced from what they can actually enable and budget for today. The rollout schedule and the end of the introductory period remain questions to resolve before treating Carbon as a production option. Google data center in The Dalles, Oregon, photographed on June 23, 2020. Archival company‑infrastructure context, not an identified Carbon training or serving facility or evidence of model access. Photo: [Lambtron / Wikimedia Commons](https://commons.wikimedia.org/wiki/File:Google_data_center.jpg), [CC BY‑SA 4.0](https://creativecommons.org/licenses/by‑sa/4.0/). Responsive crop/downscale; image adaptations retain the same license.