Gemini 4 Carbon: Release Date, Pricing, Access and vs Argon

Updated 29 October 2026. Google announced Gemini 4 Carbon on 28 October, and this guide covers who can use it now, what it will cost and how it compares with Gemini 4 Argon.

Gemini 4 Carbon is announced, but most people cannot use it yet. Google introduced it on 28 October 2026 as its strongest model, with a focus on cybersecurity and long, multi-step tasks. As of 29 October 2026, Google is opening access only to a subset of cyber defenders. Google will add more trusted partners through its Fairwind Program and then bring Carbon to developers, businesses and consumers "as soon as possible". It has not given a release date, a model ID or a list of products.

The price is already set. Carbon launches at the same introductory price as Gemini 4 Argon: $2 per million input tokens and $10 per million output tokens, with cached input 95% off. After the introductory period, $4 and $20 apply. Google has not said when Carbon's introductory period ends; for Argon, the Gemini API pricing page lists the higher rates from 1 December 2026. Below: the rollout so far, how access works, pricing, benchmarks, Carbon vs Argon, and what to do while you wait.

Related: Claude Code 5-Hour Limit Draining Fast? Agent Summaries, 8 Fixes

Gemini 4 Carbon at a glance

From Google's announcement, the Fairwind Program page and the Gemini API models page, checked on 29 October 2026:

When is the Gemini 4 Carbon release date?

There is no public release date yet. Google says that, starting this week, it will make Carbon available to a subset of cyber defenders, gradually add more trusted partners through Fairwind, and keep gathering feedback and iterating on guardrails before a wider launch. Google also says it is taking part in the U.S. government's voluntary process for pre-release model testing.

The best guide is how Gemini 4 Argon rolled out, because Carbon is following the same pattern:

Argon took 13 days from announcement to general availability. Treat that as a precedent, not a promise. Carbon's headline feature is stronger cyber capability, and Google says it is still gathering feedback and refining guardrails, so the wait could be longer. Our Gemini 4 Argon guide covers the model you can use today.

How to get access to Gemini 4 Carbon

As of 29 October 2026, there is no public waitlist, API access or app setting for Carbon. There are two routes, depending on who you are.

Cyber defense teams: the Fairwind Program

Fairwind is Google DeepMind's early-access program for defenders. The program page lists who can apply:

• Governments and national cyber authorities
• Critical infrastructure operators, such as healthcare, telecom, energy and financial networks
• Core technology platforms
• Academic labs focused on defensive benchmarking

You apply through an interest form on that page, and Google vets applicants, including background checks. Members must use phishing-resistant MFA, keep access inside their security teams and may not share or resell it. As of 29 October 2026 the page still describes Argon access; Carbon's announcement says partners will be phased in through the same program.

Everyone else: use Argon while you wait

Developers and businesses have to wait for the general release. In the meantime, gemini-4-argon is generally available in the Gemini API paid tier and Google AI Studio, and in the Gemini app for Google AI Pro and Ultra subscribers.

How much will Gemini 4 Carbon cost?

Google set the price before opening access. Per million tokens, as of 29 October 2026:

The cached prices are worked out from Google's "95% off" wording, and Artificial Analysis lists the same $0.20 at the standard rate. Google has not announced when Carbon's introductory period ends. For comparison, the Gemini API pricing page lists Argon at $2 and $10 through 30 November 2026 and $4 and $20 from 1 December 2026, with no free tier. Our Gemini 4 pricing guide has the full Argon rate card, including Batch, Flex and Priority.

Per token, Carbon will cost the same as Argon. Per task, it depends on how many tokens each model uses. Artificial Analysis measured $7.76 per task on its Intelligence Index for Carbon and $7.68 for Argon, both at the standard $4 and $20 rates.

Example (assumptions stated): a chatbot answer with 4,000 input tokens and 800 output tokens, thinking included, costs about $0.016 at the launch price, or $160 per 10,000 answers, and about $320 per 10,000 at $4 and $20. On Argon, output prices include thinking tokens, so higher thinking levels cost more. For a full build budget, see how much it costs to build an AI tool with Gemini 4.

Gemini 4 Carbon benchmarks

Google's post reports these results, evaluated without guardrails:

Artificial Analysis, which tests models independently, published more detail on 28 October 2026:

• Carbon has the top score in 4 of the 9 evaluations in the index: HLE-Diamond, Terminal-Bench-Science, AA-Omniscience and AutomationBench-AA.
• On AA-Omniscience, a knowledge and hallucination test, Carbon scores 54, ahead of Claude Opus 5.5 at 48 and Argon at 46. Its hallucination rate is 11%, up from 4% for Argon but far below the 50% measured for Claude Opus 5.5 and GPT-6 Astra.
• On AA-Briefcase, an agentic knowledge-work test, Carbon gains 126 Elo over Argon to 1609, but still trails Grok 4.7, Muse Wonder and Kimi K3.1 Pro.

Lab and vendor benchmarks are a starting point. Before you switch a production chatbot or agent, run your own test set, as our Gemini 4 vs GPT-6 vs Claude comparison explains.

Gemini 4 Carbon vs Gemini 4 Argon

Both models share the same safety stack: model training, real-time misuse defenses, prompt-injection resilience and chain-of-thought monitoring. Google also reports more than twice Argon's success rate at writing secure code and about half as many unsupported claims as Argon. It adds that Carbon reduced overreach by 26% and under-service by 29%, and that high-severity misalignment alerts in live coding environments dropped by almost 80%, without naming the baseline for those three figures.

What this means in practice:

• Agents and long tasks: Carbon is the stronger candidate once it opens, for multi-step work that calls many tools, where its agentic and prompt-injection results matter most. Our explainer on AI agents for business covers where these fit.
• Answers from your own documents: Argon remains a strong choice for an AI customer support chatbot or internal assistant. It hallucinated less on AA-Omniscience, costs the same per token, and is available today.
• Security work: Carbon's cyber results are the headline, and that is exactly where Google is gating access first.

Gemini 4 Carbon vs GPT-6 Astra and Claude Opus 5.5

Artificial Analysis ran all three on its Intelligence Index. List prices come from each provider's pricing page, as of 29 October 2026:

Carbon ranks first and hallucinates far less on this test, while GPT-6 Astra finished tasks for less money despite higher token prices. The GPT-6 Astra and Claude Opus 5.5 prices are standard API rates.

What to do while you wait for Gemini 4 Carbon

1. Build on Argon now. Carbon is a Gemini 4 model at the same token price, so prompts, tools and tests you build on gemini-4-argon should carry over. Our Gemini 4 API tutorial has Python and JavaScript examples.
2. Keep the model name in one setting. Do not guess Carbon's model ID. Read it from configuration so the switch is a one-line change when Google publishes it (example below).
3. Build a test set first. Score Argon now on 30 to 50 real tasks with known answers, then run the same set on Carbon on release day.
4. Budget at $4 and $20. Argon's rates double on 1 December 2026, and Carbon's introductory period will end too.
5. Check your data settings. Since 26 October 2026, new paid-tier Gemini API projects store Generate Content API requests by default, according to the Gemini API changelog. Existing projects keep their setting. If you send customer data, decide whether to opt out per request or in the Logs and Datasets settings.

This illustrative Python sketch uses the google-genai SDK (version 2.30.1 or later, which Google's Gemini 4 guide requires) and reads the model name from an environment variable, with Argon as the default:

# pip install -U "google-genai>=2.30.1"
# Set GEMINI_API_KEY now. Later, set GEMINI_MODEL to the Carbon model ID Google publishes.
import os

from google import genai
from google.genai import types

MODEL = os.environ.get("GEMINI_MODEL", "gemini-4-argon")

client = genai.Client()

response = client.models.generate_content(
    model=MODEL,
    contents="Summarize this support ticket in two sentences: ...",
    config=types.GenerateContentConfig(
        thinking_config=types.ThinkingConfig(thinking_level="medium"),
    ),
)
print(MODEL, response.text)

If you are planning a chatbot or agent on Gemini and want a second opinion on model choice, costs or the switch to Carbon, tell us about your project.

Gemini 4 Carbon FAQ

Is Gemini 4 Carbon available now?

Only to a small group. As of 29 October 2026, Google is opening Gemini 4 Carbon to a subset of cyber defenders, starting the week of its announcement, and plans to add more trusted partners through its Fairwind Program. Developers, businesses and consumers cannot use it yet. Google says it will bring Carbon to its products as soon as possible but has not given a date. Gemini 4 Argon, released two weeks earlier, is generally available today.

Can I use Gemini 4 Carbon in the Gemini app?

Not yet. Google's announcement does not mention the Gemini app, Google AI Pro or Google AI Ultra; it only says Carbon will reach its product surfaces after the cyber-defender phase. For comparison, Argon reached the Gemini app and AI Mode in Search for Google AI Ultra and Pro subscribers on 13 October 2026, the same day it reached the Gemini API. Watch Google's announcements for Carbon's consumer rollout.

How much will Gemini 4 Carbon cost?

Google says Carbon launches at Argon's introductory API price: $2 per million input tokens and $10 per million output tokens, with cached input 95% off. After the introductory period, the price becomes $4 per million input tokens and $20 per million output tokens. Google has not said when Carbon's introductory period ends. As of 29 October 2026, Argon's lower rates run through 30 November 2026.

What is the Gemini 4 Carbon model ID?

Google has not published one. As of 29 October 2026, the Gemini API models page lists gemini-4-argon as the only Gemini 4 model, and Carbon does not appear in the API docs or the changelog. Avoid hard-coding a guessed name. Keep the model ID in a configuration setting so you can switch from Argon to Carbon with one change once Google lists it.

What is the Fairwind Program?

Fairwind is Google DeepMind's early-access program for cyber defenders. It covers governments and national cyber authorities, critical infrastructure operators, core technology platforms and academic labs that work on defensive benchmarking. Applicants fill in an interest form and are vetted, including background checks. Members must use phishing-resistant MFA and keep access inside their security teams. Google is phasing in Carbon access for more trusted partners through this program.

Is Gemini 4 Carbon better than Argon?

On most published tests, yes. Carbon scores 59 on the Artificial Analysis Intelligence Index v5.0, 9 points above Argon, and Google reports gains in secure coding, prompt-injection resistance and long tasks. Argon still has a lower hallucination rate on AA-Omniscience, 4% against 11%, and it is available today at the same token price. Test both on your own tasks before you move production work.

Sources

• Google: Gemini 4 Carbon, our next frontier model (28 October 2026)
• Google DeepMind: Fairwind Program
• Artificial Analysis: Gemini 4 Carbon is the new #1 (28 October 2026)
• Google: Gemini 4 Argon announcement (30 September 2026)
• Google: Gemini 4 Argon general availability (13 October 2026)
• Gemini API: models
• Gemini API pricing
• Gemini API changelog
• Gemini API: Gemini 4 guide
• OpenAI API pricing
• Claude API pricing
• Hacker News: Gemini 4 Carbon discussion (28 October 2026)