Direct answer: Google announced Gemini 4 Argon on September 30, 2026, but it is not a general public release yet. Access is starting with trusted cyber defenders in Google DeepMind’s Fairwind Program. Google says broader availability will begin with paid API customers and Google AI Ultra subscribers, but it has not published a date or a public API model ID. The announced introductory API price is $2 per million input tokens and $10 per million output tokens; cached input is 95% cheaper.[1]
Gemini 4 Argon status at a glance
| Question | Verified answer |
|---|---|
| Is Gemini 4 Argon announced? | Yes — September 30, 2026 |
| Can everyone use it now? | No; the rollout starts with trusted cyber defenders |
| Who gets broader access first? | Paid API customers and Google AI Ultra subscribers |
| Public release date | Not announced |
| Public API model ID | Not announced in Google’s launch post |
| Introductory API price | $2 input / $10 output per 1M tokens |
| Price after introductory period | $4 input / $20 output per 1M tokens |
| Cached input discount | 95% off the input-token rate |
| Maximum output | Up to 1 million tokens |
Last checked: October 1, 2026. Availability and pricing can change; verify the official Google pages before purchasing or deploying.
How to check whether you have Gemini 4 Argon access
- Do not guess the model ID. Google’s announcement names the model but does not provide a public API endpoint string.
- For API access: sign in to the Google AI developer console with your paid project and inspect the available-model list. Use only the exact identifier shown by Google.
- For Google AI Ultra: open Gemini’s model selector in your signed-in account. If Argon is not listed, your account has not received access yet.
- For defensive cybersecurity access: eligible organizations can review the Fairwind Program and apply. Google says it prioritizes governments, critical-infrastructure operators and core technology platforms, and vets applicants for ethical operations.[2]
- Keep your current production model active. Do not replace a stable endpoint or promise Argon-powered features until your actual account shows access and Google publishes the required implementation details.
Important: There is no verified consumer trick, download, APK, browser extension or unofficial endpoint that unlocks Argon early. Avoid websites asking for an API key or payment to “activate” access.
Gemini 4 Argon pricing explained
Google announced an introductory rate of $2 per million input tokens and $10 per million output tokens. Cached input is priced at 95% off, which works out to $0.10 per million cached input tokens. After the introductory period, Google says standard rates will be $4 input and $20 output per million tokens.[1]
| Usage example | Intro price | Later standard price |
|---|---|---|
| 100K input + 10K output | $0.30 | $0.60 |
| 100K cached input + 10K output | $0.11 | $0.22 |
| 1M input + 100K output | $3.00 | $6.00 |
| 1M cached input + 100K output | $1.10 | $2.20 |
Examples use the announced per-token rates only. They do not include possible tool, grounding, storage, platform or enterprise charges. Google has not stated when the introductory period ends.
What the 1-million-token output limit means
Google says Argon raises the maximum output from 64,000 tokens to 1 million tokens. That is an output ceiling, not a promise that every task should generate a million tokens. The purpose is to give the model room for long, multi-step reasoning and extended work trajectories.[1]
Practical use cases may include:
- large codebase analysis, migrations and debugging;
- multi-document legal, finance and tax research;
- long-running agent workflows with many intermediate steps;
- defensive vulnerability discovery, validation and patching;
- long-video and chart analysis.
Long output can also become expensive. Set explicit completion conditions, token budgets and review gates rather than allowing an agent to run without limits.
What Google claims Argon can do
Google positions Argon as a frontier model for software engineering, enterprise knowledge work and defensive cybersecurity. In Google’s reported evaluations, Argon scored 77.9% on DeepSWE v1.1, ranked first on AutomationBench with 51.3%, reached 91.7% on LVBench and tied for first on CWE-bench v1 at 68%. These are vendor-reported benchmark results, not guarantees for your workload.[1]
Google also says the model is being tested with stronger defenses against misuse and indirect prompt injection before broader release. Fairwind access is controlled: partner organizations must use it for defensive or academic research, apply access controls and avoid malicious tasks such as creating malware.[2]
A safe pre-launch testing checklist
Teams waiting for access can prepare without writing code against an invented endpoint:
- Choose three representative tasks and save expected outputs.
- Record current model quality, latency, token use and cost.
- Remove secrets, personal data and regulated client material from test datasets.
- Define a maximum token and dollar budget per task.
- Require human approval before an agent sends, publishes, deletes or changes production data.
- Test prompt-injection resistance with untrusted webpages and documents.
- Run security-generated patches through normal code review and automated tests.
- Wait for an official model ID, SDK notes, rate limits and data-handling terms.
- Use a small canary rollout before switching production traffic.
Should you wait for Argon?
Wait if your work involves difficult coding, long professional documents, multi-step agents or authorized cyber defense and you can tolerate a phased rollout. The announced introductory pricing may make an early evaluation worthwhile.
Do not wait if you need a stable production endpoint today. Continue using an available model, build your evaluation harness now and compare Argon only after it appears in your own account with official documentation.
Frequently asked questions
Is Gemini 4 Argon available to the public?
No. Google says the initial rollout is for trusted cyber defenders through the Fairwind Program. Paid API customers and Google AI Ultra subscribers are first in line for broader access, but no general release date has been published.
What is the Gemini 4 Argon API model ID?
Google’s launch post does not publish a public model ID. Do not assume that gemini-4-argon is a working endpoint; copy the identifier from official API documentation or your signed-in model list when access arrives.
How much will Gemini 4 Argon cost?
The announced introductory price is $2 per million input tokens and $10 per million output tokens, with cached input at 95% off. Google says the later rate will be $4 input and $20 output per million tokens.
When does the introductory price end?
Google has not published an end date in the launch announcement.
Does Argon have a 1-million-token context window?
Google’s announcement specifically says the output token limit is 1 million tokens. Do not turn that into a context-window claim unless Google’s model documentation confirms it.
Can individuals apply to Fairwind?
Fairwind is aimed at approved organizations with defensive responsibilities. Google says it prioritizes governments, critical-infrastructure operators, core technology platforms and qualifying academic labs.
Sources
- Google: Gemini 4 Argon announcement — pricing, rollout, capabilities, safeguards and vendor-reported benchmarks.
- Google DeepMind: Fairwind Program — eligibility, permitted use, governance and application details.