Gemini 4 Argon is a closed API model. Google’s model id is gemini-4-argon. The announcement is dated 13 October 2026. Google does not publish weights, a parameter count, or a local VRAM number. Those fields are not published. This page does not invent them.

Price and access

Checked on 16 October 2026 against the latest-model page and the model page.

Item Published value
Input, through 30 Nov 2026 $2.00 per 1M tokens
Output, including thinking, through 30 Nov 2026 $10.00 per 1M tokens
Input from 1 Dec 2026 $4.00 per 1M tokens
Output from 1 Dec 2026 $20.00 per 1M tokens
Cached input 95% off: $0.10 per 1M, then $0.20 per 1M
Explicit caching Not supported. Implicit caching applies to prompts of at least 4,096 tokens
Free tier and Google AI Pro plan Not available, per the latest-model page
Thinking levels low, medium, high. The latest-model table lists medium as the available setting in that summary

The blog post says Argon is available to developers and enterprises through the API in AI Studio and the Gemini platform, to Google AI Ultra subscribers in Antigravity, and to Google AI Ultra and Pro subscribers in the Gemini app and AI Mode in Search. The app sentence names Pro subscribers. The API pricing sentence says the model is not on the Google AI Pro plan. Both sentences are quoted from Google. This page does not merge them into one rule.

Image generation, audio generation, and the Live API are listed as not supported on the model page. Computer use is listed as supported in preview.

Context

Limit Tokens
Unified context window 1,048,576
Max output, default synchronous 65,536
Max output with long decode 1,048,576

Inputs: text, image, video, audio, and PDF. Output: text. A parameter count is not published.

Local run

There is no public weight download on the model page, and a Hugging Face search on 16 October 2026 did not turn up an official Gemini 4 Argon checkpoint. Without weights, a VRAM total is not published and is not estimated here.

The VRAM calculator sizes supported open language models such as Qwen3.8-27B. For another closed frontier API with a published token price, see Muse Wonder. Muse Wonder’s standard rates on the same check date are $4.00 input, $0.20 cached input, and $20.00 output per 1M tokens. That is a price comparison of two published tables, not a quality ranking.

What this page does not claim

No benchmark score is copied here. No local GPU, including the RTX 4090 or DGX Spark, is a Gemini 4 Argon fit. Methodology for models this site does estimate is on the methodology page.

FAQ

Can Gemini 4 Argon run locally?

Not from a public checkpoint. Google's model page and the 13 Oct 2026 announcement describe API, AI Studio, Antigravity, and Gemini app access. They do not publish weights, a parameter count, or a VRAM requirement. This site does not estimate one.

How much does Gemini 4 Argon cost?

Google's latest-model page, updated with the Argon launch, lists introductory API pricing through 30 November 2026: $2.00 per 1M input tokens and $10.00 per 1M output tokens, including thinking tokens. From 1 December 2026 the listed rates are $4.00 input and $20.00 output. Implicit caching discounts cached input by 95 percent, to $0.10 per 1M through 30 November and $0.20 after that. Explicit caching is not supported. The same page says Argon is not on the free tier or the Google AI Pro plan.

What is the Gemini 4 Argon context length?

The model page lists a unified context window of 1,048,576 tokens. Default synchronous max output is 65,536 tokens. Long decode raises max output to 1,048,576 tokens. Inputs are text, image, video, audio, and PDF. Output is text.

When was Gemini 4 Argon released?

The Google blog post is dated 13 October 2026. It says Argon is globally available for developers and enterprises via the API in AI Studio and the Gemini platform, for Google AI Ultra subscribers in Antigravity, and for Google AI Ultra and Pro subscribers in the Gemini app and AI Mode in Search. The model page says it was last updated 14 October 2026 UTC.