Gemini 4 Argon: what is confirmed, what it costs, and how to prepare for API access.

What is Gemini 4 Argon?

Gemini 4 Argon is Google’s latest frontier model for complex, long-running work. Google says it is designed to sustain deep reasoning across real software engineering tasks, enterprise knowledge work such as legal and finance workflows, and defensive cybersecurity.

The launch is phased. Argon is first rolling out to trusted cyber defenders through Google’s Fairwind Program, with wider access planned for developers, enterprises, and consumers after further testing. That means an announcement, a published price, and a callable public endpoint are three different things right now.

Direct answers to common Gemini 4 Argon searches

These are the short answers a reader usually needs before comparing benchmarks or writing code. Checked against Google’s announcement and Google DeepMind’s model pages on 2026-10-08.

What is Gemini 4 Argon?

Gemini 4 Argon is Google’s frontier model for long-horizon coding, enterprise knowledge work, and defensive cybersecurity. Google announced it on September 30, 2026.

Is Gemini 4 Argon available yet?

Not for general public access. Google says the rollout starts with trusted cyber defenders in Fairwind; broader access is planned, but no public general-availability date is stated.

What is the Gemini 4 Argon API model ID?

No stable public model ID is confirmed in the checked Google sources. Query the models endpoint with your own key rather than guessing gemini-4-argon.

How much does Gemini 4 Argon cost?

The announced introductory price is $2 per 1M input tokens and $10 per 1M output tokens. Cached input is discounted by 95%, or $0.10 per 1M cached input tokens.

Is Gemini 4 Argon open source?

No downloadable weights are announced in the official launch material. Treat Argon as a hosted model unless Google publishes a different release.

Gemini 4 Argon benchmark comparison

The table below reproduces the comparison published on Google DeepMind’s Gemini model page. It is a vendor-reported snapshot: benchmark setup, tool access, and evaluation methodology differ, so the percentages should be read as directional rather than as a universal ranking.

Benchmark	Area	Argon	GPT-6 Astra	Claude Fable 5.1	Claude Opus 5.5
Vals Index	Knowledge work	68.9%	63.1%	65.8%	67.0%
AutomationBench	Knowledge work	51.3%	41.4%	31.4%	42.5%
DeepSWE v1.1	Agentic coding	77.9%	74.1%	67.4%	74.2%
FrontierSWE v2	Agentic coding	55.0%	65.5%	56.3%	62.3%
Vibe Code Bench	Agentic coding	91.9%	89.6%	90.3%	90.3%
GraphWalks, 256k–1M	Long context	84.2%	71.8%	65.0%	66.8%
LVBench	Multimodal	91.7%	87.5%	79.7%	83.7%
CWE-bench v1	Cybersecurity	68.0%	68.0%	58.0%	67.0%

The official page also reports 55.0% for Argon on FrontierSWE v2, where GPT-6 Astra leads this comparison. A benchmark table is useful for choosing what to test; it is not a substitute for testing the workflows that matter to you.

Cybersecurity capabilities

Google DeepMind positions Argon as a defensive cybersecurity model that can find, validate, and patch critical vulnerabilities. Its cyber page reports 85.8% on a real-world vulnerability-discovery evaluation versus 71.0% for Gemini 3.8 Flash Cyber, and 70.9% versus 58.2% on the Wiz Penetration Test Benchmark.

Those figures describe defensive evaluation results, not permission to run unreviewed autonomous changes against production systems. Keep human approval, isolated test environments, and audit logs around any security workflow.

Gemini 4 Argon API: what you can do now

There is no official public Argon model ID in the launch announcement or the public Gemini API model list at the time of this page’s update. Do not hard-code gemini-4-argon just because it matches the product name. Use the models endpoint to check what your own API key can access.

1. Keep credentials and the model name outside your code

Create a key in Google AI Studio, then set environment variables locally. Do not commit the key or paste it into a browser page.

export GEMINI_API_KEY="your-key-from-ai-studio"
export GEMINI_MODEL="the-model-id-visible-to-your-key"

2. Check whether Argon is available to your account

The model list is the source of truth for your account. An empty result means your key does not have Argon access yet.

curl -s "https://generativelanguage.googleapis.com/v1beta/models" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  | grep -i argon

3. Use the standard Gemini request shape once the ID appears

The request below keeps the model ID replaceable and caps output so a long-running job cannot accidentally consume the full announced output ceiling.

from google import genai
from google.genai import types
import os

client = genai.Client(api_key=os.environ["GEMINI_API_KEY"])
model = os.environ["GEMINI_MODEL"]

response = client.models.generate_content(
    model=model,
    contents="Review this function and list every bug you find.",
    config=types.GenerateContentConfig(max_output_tokens=8000),
)
print(response.text)

Gemini 4 Argon price calculator

Use the announced introductory rates to estimate one month of token usage. Enter values in millions of tokens. Cached input is calculated at 5% of the input rate because Google announced a 95% discount.

FAQ

Sources and update boundary

This page separates Google-reported results from our practical interpretation. Check the official pages again before making an API or procurement decision because access, prices, model IDs, and safety guidance can change during a staged rollout.

• Google Blog: Introducing Gemini 4 Argon — announcement, rollout, output limit, and introductory pricing.
• Google DeepMind: Gemini models — benchmark comparison and methodology link.
• Google DeepMind: Gemini 4 Argon cybersecurity — defensive cybersecurity evaluations.
• Gemini API pricing — public pricing and model availability reference.

Related guides

Gemma 4 vs Qwen3

Compare another open-model family when local deployment matters.

Gemma 4 Ollama setup

Use a local model when hosted frontier access is still staged.