Gemini 4 Argon Explained: Access, Cost and When to Switch

Yes, Gemini 4 is out: Gemini 4 Argon began rolling out on October 13, 2026 to Google AI Pro and Ultra subscribers in the Gemini app, with paid Gemini API access and Antigravity for Google AI Ultra. Its Standard API introductory price of $2 input / $10 output per million tokens ends November 30, 2026; budget $4/$20 from December 1. Evaluate difficult jobs now, and base a production switch on accepted outcomes at December's rate.

What Gemini 4 Argon changes for your model plan

Gemini 4 Argon, Google's model for complex, sustained work, changes which jobs deserve an evaluation before it changes your default production model. Google announced it on 30 September 2026, initially rolling it out to trusted cyber defenders and testers. The announcement describes longer reasoning, software engineering, enterprise knowledge work and defensive cybersecurity.

Since October 13, 2026, access has expanded beyond the initial defender cohort: the Gemini app rollout includes Google AI Pro and Ultra, the Gemini API offers paid access, and Antigravity offers Argon to Google AI Ultra subscribers. The practical decision is now which route fits your work and whether the result earns its December cost.

Access today: Gemini app and paid API

Google AI Pro and Google AI Ultra subscribers can use Argon in the Gemini app. Google's October 13 release notes describe a global rollout, and its subscription page explicitly lists Argon under both plans. Check that the rollout has reached your account before assigning it delivery work.

Developers can use gemini-4-argon through the Gemini API paid tier. Google's pricing page marks Argon's free tier as Not available. A consumer app subscription and metered API usage are separate access and billing routes; choose the app for interactive work and the API for an application you operate.

Antigravity and Fairwind: different access routes

Google AI Ultra subscribers can use Argon in the Antigravity desktop app and CLI, Google's coding-agent tools. The October 13 Antigravity announcement says to update to the latest version and also names eligible Gemini Enterprise subscriptions. The Pro entitlement in the Gemini app does not establish Argon access in Antigravity.

The Fairwind Program remains a controlled route for approved cybersecurity defenders, with applicant vetting, employee access restrictions and no redistribution of model access. It is no longer the eligibility test for ordinary Argon app or paid API use.

Gemini 4 Argon price: budget for December 1

Argon's Standard introductory token price matches GPT-6.1 Sol and Claude Sonnet 5.5's base rates; its December price matches Claude Opus 5.5. The introductory period ends on November 30, 2026, so a recurring workload needs the December 1 rate in its budget.

A token is a unit of text processing: you pay for input sent to a model and output it produces. The table uses USD per 1M tokens, standard direct API list prices and uncached input. Prices, access and limits were verified against the makers' own pages on 14 October 2026. Argon output pricing includes thinking tokens, the model's billed reasoning work.

Google's Argon pricing page also lists these API modes, using these exact names. Each pair below is input / output in USD per 1M tokens; none has a free tier:

• Batch: $1 / $5 through November 30, 2026; $2 / $10 from December 1, 2026.
• Flex: $1 / $5 through November 30, 2026; $2 / $10 from December 1, 2026.
• Priority: $3.60 / $18 through November 30, 2026; $7.20 / $36 from December 1, 2026.

The calculations below use Standard. Keep the processing mode consistent when comparing quotes or forecasting the price change.

OpenAI GPT-6.1 Sol was released in the API on 29 September 2026. Its quoted $2/$10 rate applies to prompts with up to 272K input tokens. Above that threshold, OpenAI doubles input and cache rates and charges 1.5 times the output rate for the full request. Large-input jobs need that qualification even when their output is short. Official OpenAI documentation supplies the threshold.

Anthropic Claude Opus 5.5 has been active since 22 September 2026. Its model page lists Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS. It gives buyers an available candidate at Argon's later base price; it does not establish equal performance on your jobs.

Anthropic Claude Sonnet 5.5 has been active since 28 September 2026, with the same listed API and cloud routes. Its model page verifies the $2/$10 base rate. It belongs in the lower-cost baseline alongside GPT-6.1 Sol when you can complete the job with either.

These are rate comparisons. Different tokenizers, output lengths, cache writes, processing tiers and tool charges can change the bill. Cloud availability also does not promise identical cloud pricing. Keep the same task and acceptance criteria when comparing costs.

For the broader plan decision, use Gemini pricing. For an available OpenAI upgrade decision, read GPT-6.1 Sol vs GPT-6 Sol.

Long decode earns its place on difficult jobs

The useful distinction is default output versus long decode, Google's mechanism for continuing a response beyond the default allowance. Google's Argon model documentation prints 65,536 tokens for default synchronous output and up to 1,048,576 with long decode. That larger figure is constrained by a unified window shared across input, thinking and output, rather than an additional output budget on top of your input.

For the Interactions API, Google's long-decode guide requires background execution and a max_output_tokens setting above 65,536 to go beyond the default. The available space is 1,048,576 minus your input tokens. Those are documented API conditions, not a promise that every chat or coding-agent session can produce that much.

Input context is what the model can read at once. Output allowance is how much it can produce. A giant reading window cannot help a job that runs out of room while working through a solution, and a giant output allowance does not guarantee a correct solution.

The extra allowance matters when a difficult response runs out of room for reasoning and output. A code migration can require inspecting files, changing interfaces, checking tests and revising a proposed patch. A financial investigation can require reconciling documents, following discrepancies and assembling a supported conclusion. Ending early can leave the expensive part unfinished.

Google names codebase migrations, algorithm design, multi-step financial research, legal research and drafting, chart analysis and long-video understanding as target areas. The jobs worth evaluating have a clear completion condition:

• A solo technical builder: try a repository migration that currently ends with an incomplete patch. Judge the result by tests and reviewable changes, rather than the length of its explanation.
• A mid-market CTO: choose a bounded maintenance job with an owner and a rollback path. Ask whether the model reduces restarts and reviewer work before considering a larger rewrite.
• A senior operator: choose a research packet that requires reconciling several documents and checking charts. Require traceable evidence for the final recommendation.

Those are proposed evaluations, not results from tests performed for this article. Google's internal migration examples also retain automated checks and human review before production. A multi-step agent whose individual turns are short does not automatically need long decode; Google's guide makes that distinction explicit.

A funded founder with a reliable short-request workflow has little reason to pay for unused headroom. The Gemini review addresses the broader product and workflow fit; Argon's larger allowance only matters if it removes a constraint in a named job.

The quarterly budget and the cost per accepted job

An unchanged uncached Standard workload costs twice as much from December 1, 2026. Both input and output rates double on that date. A forecast spanning the change needs separate usage totals for the period through November 30 and the period from December 1, using Google's dated prices.

Consider a hypothetical workload of 100 jobs per month, each using 100K billed input tokens and 20K billed output tokens, including thinking. That totals 10M input and 2M output tokens:

• Standard through November 30: 10 × $2 + 2 × $10 = $40 per month.
• Standard from December 1: 10 × $4 + 2 × $20 = $80 per month.

For three full months of that hypothetical usage from November 2026 through January 2027, the token-only budget is $200: $40 for November plus $80 each for December and January. Use your expected billed token totals for each period; this is an illustration, not a usage forecast.

At identical billed quantities, GPT-6.1 Sol and Claude Sonnet 5.5 also total $40 at their quoted base rates; Claude Opus 5.5 totals $80. The example assumes each request stays within Sol's base input threshold. It does not assume all models tokenize the same material or consume the same amount of output.

Cached input reduces one part of the bill

Google's pricing page lists Standard cached input at $0.10 per 1M tokens through November 30, 2026, then $0.20 from December 1. That is 95% below uncached input in either period. Argon supports implicit caching only, where the service recognizes reusable input; explicit caching is not supported.

If the hypothetical workload achieves 90% input cache hits, its input becomes 1M uncached and 9M cached tokens. Keeping output unchanged:

1 × $2 + 9 × $0.10 + 2 × $10 = $22.90 during introduction.

That is a token-rate illustration, excluding other charges. A 95% cached-input discount does not make the entire job 95% cheaper, because output still contributes $20 in this example. A reusable repository brief may help input costs; long reasoning still needs its own output budget.

At the same assumed cache-hit rate, the December bill becomes 1 × $4 + 9 × $0.20 + 2 × $20 = $45.80. Treat the cache-hit rate as a measured workload property, not a guaranteed discount on all input. Google's long-decode documentation also bills repeated input across continuation steps, so use reported usage when estimating those runs.

The crossover is accepted work

At the example's later rate, Argon costs $0.80 per attempt, against $0.40 at a $2/$10 base rate. If billed tokens per attempt stay fixed, doubling the fraction of accepted attempts offsets doubled token cost.

For illustration, a baseline accepting 40% of attempts costs $0.40 ÷ 0.40 = $1 per accepted job. A model accepting 80% at $0.80 also costs $1 per accepted job. These acceptance rates are assumptions used to explain the crossover, not scores measured for Argon or a competitor.

If your baseline already accepts 60%, acceptance alone would need to reach 120% to erase that premium under the same assumptions. It cannot. The business case then needs lower token consumption, less human review, or valuable work your current model cannot finish.

Output volume also has a bill attached. A hypothetical 1M billed output tokens across your workload costs $10 through November 30 or $20 from December 1 at Standard rates, before input and other charges. That is output-only arithmetic, not the total bill for a long-decode run.

Google's benchmarks identify what to evaluate

Google's published numbers support a targeted shortlist of jobs. They do not establish your application's acceptance rate or savings. No hands-on model tests were run for this piece.

All four scores below are Google's reported results from its 30 September announcement.

The percentages measure different things under different evaluation conditions. Do not average them into a capability score or substitute DeepSWE's result for your own code acceptance rate.

In particular, long-video understanding does not prove reliable software changes. A vulnerability-remediation result does not promise the same cyber capabilities or safeguards for every customer. Pick the benchmark closest to your blocked job, then verify the outcome under your own constraints.

What builders, operators and buyers should do differently

Builders should act now if their paid API or Antigravity access lets them evaluate a job their current model repeatedly fails to finish. Start with a bounded task and compare accepted outcomes, total billed tokens and reviewer effort. Keep an available provider as the baseline.

Builders whose account has not received the rollout should prepare that evaluation and continue shipping. For a solo builder producing short patches, or a founder whose existing workflow is dependable, the larger output ceiling changes little until a recurring job needs it.

Operators should prepare the checks that make a long run reviewable. A CTO evaluating a migration needs clear scope, regression tests and an accountable reviewer. A senior operator evaluating research needs a source set and a way to verify the conclusion. Longer execution increases the amount of work you must inspect.

Buyers should reserve budget at the December rate and make migration conditional. Access through the intended product, a finished result and a better cost per accepted outcome are the conditions. A published leaderboard or a consumer subscription purchase does not satisfy them.

Wait when the job already completes reliably at a lower price, when access is unconfirmed, or when you cannot measure whether the result is acceptable.

What's overhyped: the shortcuts to avoid

App access is not the same entitlement as API or Antigravity access. Google AI Pro and Ultra cover Argon in the Gemini app; the Gemini API requires paid-tier access, and the Antigravity announcement names Google AI Ultra and eligible enterprise subscriptions. Keep the product and entitlement together in any procurement note.

Long decode is not a reliability score. More room can allow a model to finish difficult reasoning; it can also produce more material to review. Longer work earns its cost when it improves an accepted result.

The Arena “Argon checkpoint” story is unverified for this decision. Google's announcement does not authenticate that pre-release checkpoint identity. Do not carry leaked model names, specifications or benchmark tables into a production plan. Use the named model and its current official documentation.

The introductory price is not a durable savings case. A Standard business case that works at $2/$10 and fails at $4/$20 stops working on December 1, 2026. The November 30 expiry is a budgeting deadline. Calculate both cases before committing.

The missing evidence that matters is your cost per accepted job at the December price. Google's published results are a reason to evaluate that question.

The Monday move: prepare a decision your team can measure

Give one owner a concrete evaluation brief next week. The result should be a switch decision you can defend before the introductory price ends.

1. 1

   Name the blocked job

   Choose a recurring task such as an incomplete repository migration, an authorized vulnerability fix or a document-and-chart research packet. Write down what an acceptable finished result looks like.

2. 2

   Record the baseline

   Use your current available model. Track billed input and output, retries, acceptance and reviewer time. Keep the task and acceptance standard consistent.

3. 3

   Prepare the later-rate budget

   Price the Standard workload at $4/$20 from December 1, 2026. Keep the $2/$10 case separately through November 30. Confirm access in the product you intend to use before assigning delivery work to Argon.

4. 4

   Switch only when the job improves

   With your chosen access route, compare completed acceptable work and total effort. Expand use when the gain covers the token premium and any extra review; keep the baseline when it does not.

Gemini 4 Argon questions

Is Gemini 4 Argon out?

Yes. Since October 13, 2026, Argon has been rolling out to Google AI Pro and Ultra subscribers in the Gemini app. Developers can use the Gemini API paid tier, with no Argon free tier. Google AI Ultra subscribers can also use Argon in Antigravity desktop and CLI; eligible enterprise subscriptions have a separate route. See the app release notes and Antigravity announcement.

When can we expect Gemini 4?

The wider rollout began October 13, 2026. Google AI Pro and Ultra are the Gemini app routes, paid-tier access is the Gemini API route, and Google AI Ultra is the consumer subscription named for Argon in Antigravity. Check your account as the rollout reaches users.

What is Gemini not allowed to do?

For Argon through Fairwind, Google's program limits use to permitted defensive and research purposes, bans malicious activity and prohibits sharing, redistributing or selling access. These are Fairwind-specific rules; the public app and paid API rollout does not grant Fairwind privileges.

What is Gemini 4?

Gemini 4 Argon is Google's model for complex, sustained work across software engineering, enterprise knowledge tasks and cyber defense. Google's model docs distinguish default synchronous output of 65,536 tokens from long decode within a shared 1,048,576-token input, thinking and output window.

When did Gemini 4 launch?

Google announced Gemini 4 Argon and its restricted defender rollout on September 30, 2026. The wider Gemini app and Antigravity rollout began October 13, 2026, and Argon is now listed on the Gemini API paid tier.

What is the disadvantage of using Gemini?

For an Argon adoption decision, the immediate tradeoffs are paid access, Standard token prices doubling on December 1, 2026, and the need to prove value on your own workload. Long reasoning and generated output also create more work to verify.

How much does Gemini cost per month?

Argon's API bill depends on usage. Standard pricing is $2/$10 per million input/output tokens through November 30, 2026, then $4/$20 from December 1. The example in this article costs $40 per month before that change and $80 after it. Google AI Pro and Ultra are separate app subscriptions; consult the subscription page for your local monthly price.

What are the risks of using Gemini?

For long Argon workflows, review mistaken code changes, unsupported conclusions and unintended tool actions. Google describes defenses against misuse, prompt injection and misalignment, alongside hardened sandboxes. Those measures still leave your organization responsible for checking accepted work.

Is Gemini really worth it?

Evaluate Argon when a specific difficult job needs its capabilities and the December-rate cost per accepted result can improve. Keep your current production choice when it already completes the job reliably, or when you cannot establish a better accepted outcome.

For practical model and workflow decisions as access changes, subscribe to the newsletter.