Google announced Gemini 4 Argon on 30 September, but only a few security testers could use it. On 13 October it opened to everyone: paying API customers, Google AI Pro and Ultra subscribers, and Google's coding tool Antigravity. Three facts matter more than the headline. The price is an introductory one that doubles on 1 December. The only independent score puts Argon level with OpenAI's flagship, not clearly ahead. And you no longer need the API to try it: it is included in Google's AI Pro plan. This article was first written at the announcement and rewritten on 14 October now that you can actually buy it.
What changed on 13 October, and who can use Argon today
Google DeepMind calls Argon a model "built to sustain deep reasoning across complex, long-horizon workflows". In plain terms, it is designed for long jobs with many steps, not quick chat answers. Google aims it at software engineering, "enterprise knowledge work like legal and finance", and cybersecurity defence.
- Developers and businesses. Google says Argon is "globally available today for developers and enterprises via the API" in Google AI Studio and the Gemini platform. It is a paid model, with no free tier.
- Subscribers. Google AI Pro and Ultra subscribers get it in the Gemini app and in AI Mode in Search. Google Antigravity, the agentic coding tool, is for Ultra subscribers.
- Still restricted. The version without the usual cyber guardrails stays with a small group of trusted defenders in Google's Fairwind Program.
- What is new under the hood. Argon can write up to 1 million tokens in a single answer, up from 64,000 on Google's previous models. A token is roughly three quarters of a word.
The price: $2 and $10 until 30 November, then $4 and $20
Google prices Argon at $2 per million input tokens and $10 per million output tokens, with cached input at 95% off. The introductory period ends on 30 November 2026. From 1 December the price is $4 in and $20 out, which VentureBeat also reports. Here is how that compares with the models you can buy, per million tokens:
| Model | Input | Output |
|---|---|---|
| Gemini 4 Argon (until 30 Nov) | $2.00 | $10.00 |
| Gemini 4 Argon (from 1 Dec) | $4.00 | $20.00 |
| Claude Opus 5.5 | $4.00 | $20.00 |
| GPT-6 Astra | $10.00 | $50.00 |
| Claude Sonnet 5.5 | $2.00 | $10.00 |
| GPT-6.1 Sol | $2.00 | $10.00 |
| Gemini 3.8 Flash | $0.75 | $3.75 |
For now, Google is selling its flagship at the price of the mid-tier workhorse models. At full price it lands exactly on Claude Opus 5.5, and still well below GPT-6 Astra. Competitor prices are as listed by Artificial Analysis on 14 October.
Budget for the price in the footnote, not the headline: Argon is a $4-in, $20-out model being sold at half price for seven weeks.
The subscription route: Argon is in Google AI Pro, not just the API
You do not have to write code to use Argon. Google's plans page lists Gemini 4 Argon as included in Google AI Pro and in Google AI Ultra, which adds higher usage limits, Antigravity and Project Genie. Plan prices differ by country, so check the figure Google shows for your region before you decide.
- For one person or a small team using the chat app. AI Pro is the cheapest way to try Argon. You get it in the Gemini app and in Search, with usage limits that Google sets and can change.
- For developers. Google's announcement puts Antigravity, its agentic coding tool, with Ultra subscribers. For your own software, the API is the direct route.
- For automation. A chat subscription cannot run your invoices or customer replies overnight. That needs the API, and the API is billed per token, as in the table above.
Google's own benchmarks: where Argon wins, and where it does not
Google DeepMind publishes a comparison table on its Gemini model page. Every number below is Google's, from Google's own runs, not an independent check. By our count of that table, Argon has the top score on 13 of its 19 rows, ties on one, and loses on five. The wins are concentrated in exactly the kind of work an owner or finance team would hand over. Google's announcement also says Argon ranks first on the Vals Index, which covers finance, coding, legal and tax work.
| Test (Google's numbers) | Argon | GPT-6 Astra | Opus 5.5 |
|---|---|---|---|
| Vals Index (finance, coding, legal, tax) | 68.9% | 63.1% | 67.0% |
| Vals Finance Agent v2 | 65.4% | 53.5% | 58.6% |
| Harvey Legal Agent Benchmark | 19.6% | 5.4% | 3.8% |
| AutomationBench (Zapier) | 51.3% | 41.4% | 42.5% |
| DeepSWE v1.1 (coding) | 77.9% | 74.1% | 74.2% |
| Chartography (reading charts) | 71.6% | 71.0% | 66.3% |
| FrontierSWE v2 (coding) | 55.0% | 65.5% | 62.3% |
| Terminal-bench 4.0 | 57.4% | 58.2% | 66.4% |
| OSWorld-2.0 (computer use) | 69.2% | 72.6% | not listed |
What the wins mean
- Finance and legal work. Vals Finance Agent tests multi-step financial research. Harvey's benchmark tests legal research and drafting. Argon's legal score is low in absolute terms, but nearly three times the next best in Google's table (Claude Fable 5.1, at 6.7%). Google is clearly aiming at the back office.
- Office automation. AutomationBench, built by Zapier, measures whether a model can carry a business task through from start to finish. Argon scores 51.3%, nearly nine points ahead of Opus 5.5. That is a real lead, but a score of about half still leaves a lot of room.
- Charts, documents and video. Google says Argon is state of the art on LVBench (long video understanding) at 91.7%, and leads on reading charts. That matters if your work lives in scanned statements, reports and dashboards.
What the losses mean
- Operating a computer. Argon trails GPT-6 Astra on OSWorld-2.0 and trails both rivals on Terminal-bench 4.0. If you are building an agent that clicks through websites or runs commands, Google's own table says Argon is not the leader.
- Some coding. It wins DeepSWE but loses FrontierSWE v2 by ten points to Astra. Coding results depend heavily on the test. Run your own before you move a development team.
The independent number: level with GPT-6 Astra, not ahead
Vendor benchmarks are chosen by the vendor. Artificial Analysis is an independent company that runs every major model through the same set of tests and publishes two numbers: an Intelligence Index score, and the average cost of completing its test tasks at the model's real prices. Its Argon page, checked on 14 October, shows:
- Score. Argon (at high effort) scores 53, ranked 8th of 216 models. GPT-6 Astra at max effort also scores 53. Claude Opus 5.5 scores 58 at max effort and 54 at high effort.
- Cost per task. $1.99 at the introductory price. Argon used 110 million output tokens to run the index, which is why a $10 output price does not make it the cheapest model at its score.
- Speed. Still not published, so we cannot yet say how fast it answers.
Here is what it costs to reach about the same level of capability with each model, from the Artificial Analysis leaderboard on 14 October:
Read that chart carefully, because it changes the story. Until 30 November, Argon is good value for a flagship: it matches GPT-6 Astra's score for well under Astra's cost. But Claude Opus 5.5 at high effort scores a point higher for a little less money. GPT-6.1 Sol gets within one point for about a third of the cost. And from 1 December, Argon costs more per task than every model on this chart except Muse Wonder.
Argon against Claude Sonnet 5.5
Sonnet 5.5 is the closest match on price, because both list at $2 in and $10 out until Argon's introductory price ends. Artificial Analysis tests Sonnet at several effort settings, and the answer depends on which one you use:
- Same score, same cost today. Sonnet 5.5 at xhigh effort scores 52 at $2.01 per task. Argon scores 53 at $1.99. That is a tie until 30 November.
- After 1 December, Sonnet is about half the price. Argon's price doubles to about $3.98 per task for the same work, while Sonnet stays at $2.01, assuming Anthropic does not change its prices.
- Sonnet wins on top score. At max effort, Sonnet 5.5 scores 56, three points above Argon, at $5.46 per task. It also used 420 million output tokens to run the index, the most of any model Artificial Analysis lists, so you pay for that effort.
- Argon wins where Google says it does. Google's table has no Sonnet column, so there is no vendor comparison on finance, legal or document work. Only your own test will answer that.
Two caveats in Argon's favour. First, a general index averages many kinds of work. Argon's lead in Google's table is in finance, legal and document work, so it could do better on your jobs than its average suggests. Second, testers re-run models after launch, so treat the figures above as a snapshot from 14 October.
A million-token answer is a feature and a budget risk
The change from a 64,000-token output limit to 1 million tokens sounds technical. It is really about money. Output is the expensive side of the price list, and a model allowed to write a million tokens can, in principle, bill you for a million tokens in a single answer.
- The arithmetic. One million output tokens costs $10 at the introductory price and $20 at the full price. That is a single request, before any input, retries or follow-up steps.
- Why it happens in practice. Long agent runs, "think as long as you need" settings and loops that keep asking for more are exactly how a fixed-price task turns into a surprise bill.
- The fix is boring and it works. Set a maximum output length on every request, set a monthly spending cap in the provider's console, and log the cost of each job, not just the monthly total.
Should you switch? A decision by situation
- Already shipping on GPT-6.1 Sol, Sonnet 5.5 or Opus 5.5? Stay. On independent numbers they are within a point or two of Argon, Sol is far cheaper per task, and Sonnet 5.5 at xhigh effort is about half Argon's price after 1 December. Switching later is a configuration change if you built for it, and a rewrite if you did not.
- Document, finance or legal work? Test Argon this month. That is where Google claims its biggest lead, and the introductory price makes a trial cheap. Twenty real tasks with known good answers will tell you more than any benchmark table: reconcile a bank statement, summarise a contract, read five supplier invoices. Run the same twenty on your current model and compare quality and cost per job.
- Agents that operate a computer? Wait. Google's own table has Argon behind Astra and Opus on the computer-use and terminal tests.
- Planning a budget? Use $4 and $20. If a workflow only makes financial sense at the introductory price, it does not make sense.
- Just want to try it? AI Pro is the cheap door. It is a fair experiment for one person. Do not buy Ultra just to get Argon.
Where this leaves the Gemini, ChatGPT and Claude choice
Argon is Google's answer at the top end. On Google's numbers it leads most tests, and on the one independent score it has drawn level with OpenAI's flagship. That makes this a three-way race at the top again, which is good news for anyone paying the bills. For most small and medium businesses the rule has not changed: the cheapest model that does your job well is the right one, and independent cost per task is the number to choose by.
We compared the two newest mid-tier models in Claude Sonnet 5.5 vs GPT-6.1 Sol, the subscription prices in rupees in ChatGPT vs Claude vs Gemini plans and prices, and the newest challenger in Meta Muse Wonder. For API prices across every major vendor, including the Gemini Flash rise coming in January, see AI API pricing for 2027.
If you want the benefit of these models without managing them yourself, that is what our products are for. The AI Chat Assistant answers customer questions on your website and hands complex chats to your team on WhatsApp. IO Snack Accounts uses AI to categorise entries and flag anomalies as you post them, which is the kind of finance work Argon is being aimed at. If you are planning something bigger and want a second opinion on which model fits it, talk to us.
