In this guide · 8 sections
- 01What Google launched on 13 October 2026
- 02Where you can use Gemini 4 Argon
- 03Gemini 4 Argon API price, before and after 1 December 2026
- 04How Argon ranks on published benchmarks
- 05Who should switch to Argon and who should wait
- 06How to run your own 30-minute test
- 07Do it yourself or get help?
- 08Frequently asked questions
Short answer
Gemini 4 Argon is Google's flagship model, launched 13 October 2026. It costs US$2 / US$10 per million input / output tokens until 30 November 2026, then US$4 / US$20. It ranks near the top of published leaderboards but not first on every one. Most small businesses can wait.
What Google launched on 13 October 2026
Google's launch post describes Argon as its frontier model for "deep work" on long, multi-step tasks in coding, finance, legal, cybersecurity and science. It is the first model in the Gemini 4 generation. Google says it ranks first on the Vals Index, which scores models on finance, coding, legal and tax work.
The Google Cloud announcement adds two claims that matter for business use: Argon is trained to pause before irreversible steps and say when it could not finish a task, and Google says it is its most resistant model yet to indirect prompt injection (hidden instructions in emails or web pages). These are vendor claims, not independent tests.
Where you can use Gemini 4 Argon
Access depends on which Google product you pay for. As of 20 October 2026, Google's posts list these routes:
| Where | Who gets Argon | Notes |
|---|---|---|
| Gemini API (AI Studio) | Paid API projects | Model ID gemini-4-argon; not on the free tier |
| Gemini Enterprise Agent Platform | Google Cloud customers | GA since 13 Oct 2026; global, US and EU regions |
| Gemini for business (Gemini Enterprise) | Gemini Enterprise customers | Admin toggle off by default; switches on by default from 27 Oct 2026 |
| Gemini app | Google AI Pro and Ultra subscribers | Personal subscriptions |
| Google Antigravity | Google AI Ultra subscribers | Google's coding tool |
The launch posts do not say Argon is in the Gemini app for every Google Workspace Business plan. If your staff use Gemini through a Workspace account, open the model picker in the Gemini app to see what your edition offers. For what Workspace includes today, see our comparison of Copilot vs Gemini for small business.
Gemini 4 Argon API price, before and after 1 December 2026
Google's Gemini API pricing page lists an introductory price that ends on 30 November 2026. Prices are in US dollars per million tokens, as of 20 October 2026.
| Tier | Input to 30 Nov | Output to 30 Nov | From 1 Dec 2026 (in / out) |
|---|---|---|---|
| Standard | $2.00 | $10.00 | $4.00 / $20.00 |
| Batch and Flex | $1.00 | $5.00 | $2.00 / $10.00 |
| Priority | $3.60 | $18.00 | $7.20 / $36.00 |
| Cached input (standard) | $0.10 | n/a | $0.20 |
At the standard rate from December, Argon costs the same per token as Claude Opus 5.5 and less than GPT-6 Astra, according to our AI API pricing comparison. Thinking tokens count as output, so long reasoning answers cost more than their visible length suggests.
How Argon ranks on published benchmarks
We have not run our own tests. These are published scores with their dates. On the Artificial Analysis LLM leaderboard checked on 20 October 2026, Gemini 4 Argon (high) has an Intelligence Index of 53 at about US$1.99 per task. The top entries are Claude Opus 5.5 (max) at 58 and Claude Sonnet 5.5 (max) at 56; GPT-6 Astra (max) also scores 53, at US$3.26 per task.
- Google reports Argon first on the Vals Index (finance, coding, legal, tax).
- Google says it ties for first on CWE-Bench v1 at 68% for fixing security vulnerabilities.
- Google says it leads the Gray Swan indirect prompt injection benchmark, with 0.7% attack success.
Leaderboards move every few weeks, and different benchmarks favour different models. Our guide to LLM benchmarks explained shows how to read them without over-trusting a single number.
Who should switch to Argon and who should wait
For most small businesses the model is not the bottleneck. The bigger questions are whether staff use AI at all and whether your data is set up for it.
Try Argon now if
- You already build on the Gemini API and need better long-document or multi-step results
- You review long contracts, tenders or financial packs and pay for Gemini Enterprise
- You can test during the introductory price window
Wait or skip if
- Your chatbot answers FAQs; a Flash model is far cheaper
- You use ChatGPT or Claude and they work well
- You only have a Workspace Business plan and Argon is not in your model picker
For a wider comparison of the three main assistants, see ChatGPT vs Claude vs Gemini for business. For personal plans, our guide to Google AI Pro vs ChatGPT Plus vs Claude Pro lists Singapore prices.
How to run your own 30-minute test
Published scores tell you little about your own documents. A short side-by-side test does.
Pick three real tasks
For example: summarise a 20-page supplier contract, draft a reply to a difficult customer email, and check a spreadsheet of invoices for errors. Remove personal data first.
Run each task in two models
Use Argon in AI Studio (paid project) or the Gemini app, and the model you use today. Give both the same prompt.
Score the answers
Mark each on accuracy, whether it followed your instructions, and how much editing it needed. Note any made-up facts.
Check the cost
AI Studio shows token counts. Multiply by the December price to see what a month of that task would cost.
Decide per task, not per brand
Many businesses end up using a cheap model for routine replies and a flagship like Argon only for the hard 10%.
Do it yourself or get help?
Trying Argon in the Gemini app or AI Studio takes minutes, and the test above is enough to decide for most offices.
Before you switch
If you want a website assistant or an internal tool built on whichever model fits, our AI Website Assistant package (S$790, paid upfront) sets one up on your own content, and Workflow Automation (S$390) connects AI steps to your email, forms and spreadsheets. IT Sorted is a Singapore-based team of developers with over 10 years of experience.
Frequently asked questions
Google made Gemini 4 Argon generally available on 13 October 2026, in the Gemini API, Google AI Studio, the Gemini Enterprise Agent Platform, Gemini for business and the Gemini app for Google AI Pro and Ultra subscribers. Before that it was available only to selected testing partners, including cyber defenders in Google's Fairwind Program.
On Google's pricing page as of October 2026, the standard rate is US$2 per million input tokens and US$10 per million output tokens through 30 November 2026. From 1 December 2026 it rises to US$4 and US$20. Batch and Flex are half the standard rate. Argon is not available on the free tier.
It depends on the benchmark and the task. Google reports Argon first on the Vals Index. On Artificial Analysis's leaderboard on 20 October 2026, Argon scored 53, level with GPT-6 Astra and behind Claude Opus 5.5 at 58, at a lower cost per task. Test it on your own work before switching.
Google's launch posts name Gemini for business (Gemini Enterprise) and personal Google AI Pro and Ultra subscriptions, not every Workspace Business edition. Open the Gemini app with your work account and check the model picker. Your Workspace admin can also confirm which Gemini features your edition includes.
On the paid Gemini API, Google's pricing page says content is not used to improve its products. On Google Cloud, Argon supports data residency and customer-managed keys in US and EU regions. Personal Gemini app subscriptions have different terms, so do not paste customer data into a personal account.