Gemini 4 Argon Is Half Price Until November 30. What Changes When the Price Doubles, and What to Check Now

Google made Gemini 4 Argon available to enterprise developers this week, with a pricing detail that matters if you build on it: the current price is introductory, and it doubles after November 30, 2026.

Here is what Google documents, what the claims in its launch post are worth, and a short way to plan for the price change. Everything below comes from Google's own pages, so treat performance statements as vendor claims.

What Google says

In its Google Cloud announcement, dated October 13, Google describes Argon as its new frontier model for long, multi-step work. It says that starting today it is available in Gemini for business on web, desktop and mobile, and through the API in the Gemini platform under the model ID gemini-4-argon, in Google AI Studio, and in Google Antigravity.

Google says it trained Argon to respect system boundaries, report negative findings honestly, and pause for confirmation before irreversible steps.

The price

Per the announcement, Argon launches at an introductory rate of $2 per million input tokens and $10 per million output tokens through November 30, 2026. Cached input tokens are priced at 95% off. After November 30, standard pricing of $4 input and $20 output per million tokens applies.

The Gemini API documentation gives the same figures. In plain terms, a workload costs about twice as much in December as in November, if usage stays the same.

What the API docs add

According to the same documentation:

• The model ID is gemini-4-argon, with a default thinking level of medium and tunable levels low, medium and high.
• Input, thinking and output share one window of 1,048,576 tokens.
• A single response can go past 64K tokens, up to 1M tokens minus the input, using background execution. The model generates in hops of up to 64K tokens, and Google's servers run the continuation hops.
• Agentic video understanding is generally available, and the Antigravity agent in Gemini Managed Agents supports Argon.

One practical point from the docs: max_output_tokens is a hard limit that covers thinking and output together, so set it high enough that it does not cut an answer short.

Claims in the launch post, labeled

These are Google's statements, and I have not verified them:

• Argon leads the Vals Index and Harvey's Legal Agent Benchmark.
• It ties for first on CWE-Bench v1 with 68% on remediating security vulnerabilities.
• It leads the Gray Swan Indirect Prompt Injection benchmark, cutting attack success across 15 adaptive attempts to 0.7%.
• Google's own engineering teams report results, such as optimizations in its open-source ADK that cut in-memory session prep time by 40% in long sessions.

The post also includes customer quotes from partners. Those are testimonials chosen by the vendor. Google also says it is releasing a version without cyber guardrails to vetted defenders through its Fairwind Program.

How I would plan for the price change

These are suggestions, not tested results.

1. Budget at the December price. If your workload is only affordable at $2 and $10, it is not affordable yet.

2. Measure tokens per finished task now. The introductory window is a cheap time to collect real numbers on your own tasks.

3. Compare thinking levels. The default is medium. Check whether low is good enough for simple steps, since thinking tokens count toward output.

4. Use caching where it fits. Google lists cached input at 95% off, so repeated long context is the first place to look.

5. Set output limits deliberately. Long decode can produce very long responses, which means a large output bill if unchecked.

6. Set a date to revisit. Put November 30 in your calendar and re-check Google's pricing page, since terms can change.

What I would not conclude

Do not treat vendor benchmark leadership as a result on your workload. Do not assume the December price is final either; I only know what Google has published today.

Frequently asked questions

What is the Gemini 4 Argon model ID?

gemini-4-argon, per Google's API documentation.

What does it cost?

$2 per million input tokens and $10 per million output tokens through November 30, 2026, then $4 and $20.

Where is it available?

Per Google: Gemini for business, the Gemini API, Google AI Studio and Google Antigravity.

What is the context window?

A unified 1,048,576-token window shared by input, thinking and output.

Can a response be longer than 64K tokens?

Yes, up to 1M tokens minus input, using background execution, according to the docs.

Are the benchmark claims independent?

No. They come from Google's launch post.

What should I do before November 30?

Measure cost per finished task, test thinking levels, and plan for the doubled price.

Sources

• Google Cloud Blog, Gemini 4 Argon available on Google Cloud: availability, pricing, benchmark and customer claims (October 13, 2026).
• Gemini API docs, What's new in Gemini 4 Argon: model ID, pricing, context window, long decode, thinking levels.
• TLDR AI, October 14 issue: discovery pointer only.