Gemini 4 Argon

Key facts

Context window

1,048,576

tokens

Max output

65,536

tokens, default

Input price

$2.00

per 1M tokens, standard

Output price

$10.00

per 1M tokens, standard

Extended output (Long decode)

1,048,576

tokens, special mode only

Released

Oct 13, 2026

General availability date from Google's blog post; Google announced the model on Sep 30, 2026 to a limited set of trusted cyber defenders.

Knowledge cutoff

Mar 2026

API model ID

gemini-4-argon

• Long decode: Default synchronous output is 65,536 tokens. Long decode generates longer responses in 64K-token hops, needs the Interactions API with background:true, and shares the 1,048,576-token window with input and thinking. It is billed at the normal token rates for every hop.
• Google describes the 1,048,576-token limit as a unified input and output token limit, and does not state a separate maximum input.

About Gemini 4 Argon

Gemini 4 Argon is Google's flagship reasoning model, built for complex software engineering, deep mathematical and scientific problem solving, autonomous multi-step agents and long decode up to 1M tokens. It takes text, image, video, audio and PDF input and returns text. Google lists introductory pricing of $2 input and $10 output per million tokens, rising to $4 and $20 on December 1, 2026.

• Output prices include thinking tokens.
• Flex inference is priced the same as Batch.
• Search grounding is billed separately: 5,000 requests per month free (shared across Gemini 3 and newer models), then $14 per 1,000 requests, capped at 1,500 requests per day.
• Not available on the free tier.
• Google lists no prompt-length threshold tier for this model.
• Explicit caching is not supported, so no cache storage price is listed.

Input modalities

Text, Image, Video, Audio, PDF

Output modalities

Text

• ReasoningReasoning: ✓ Yes

  Thinking levels: low, medium, high.

• Tool / function callingTool / function calling: ✓ Yes
• Structured outputsStructured outputs: ✓ Yes
• Web searchWeb search: ✓ Yes · API tool

  Grounding with Google Search.

• Code executionCode execution: ✓ Yes · API tool
• Computer useComputer use: ✓ Yes · API tool
• Image generationImage generation: ✕ No

  Google lists image generation as not supported for this model.

• Prompt cachingPrompt caching: ✓ Yes

  Implicit caching only; explicit caching is not supported.

• Batch APIBatch API: ✓ Yes

API vs. apps. These limits and prices describe the Gemini API model. The Gemini app and Google AI plans can apply different limits and are billed through subscriptions, not per token.

• Google lists introductory pricing that doubles on December 1, 2026; the current price shown depends on the date.
• Agentic video understanding defaults to static.
• Google's model card gives a March 2026 knowledge cutoff and says knowledge may be less up to date in some domains.

1. Gemini API docs — Gemini 4 Argon

   Supports: status, context window, max output, modalities, capabilities, pricing

   OfficialChecked Oct 31, 2026

2. Gemini API — Models

   Supports: status

   OfficialChecked Oct 31, 2026

3. Gemini API — Pricing

   Supports: pricing

   OfficialChecked Oct 31, 2026

4. Gemini API — Deprecations

   OfficialChecked Oct 31, 2026

5. Google — Gemini 4 Argon is now available (Oct 13, 2026)

   Supports: release date

   OfficialChecked Oct 31, 2026

6. Google — Gemini 4 Argon announcement (Sep 30, 2026)

   Supports: benchmarks

   Google blog post; scores are Google-reported, stated in the post text.

   Vendor benchmarkChecked Oct 31, 2026

7. Google DeepMind model card — Gemini 4 Argon

   Supports: knowledge cutoff

   OfficialChecked Oct 31, 2026

• Gemini 3.1 Pro (Preview)Compare with Gemini 4 Argon
• Gemini 3.5 Flash-LiteCompare with Gemini 4 Argon
• Gemini 3.6 FlashCompare with Gemini 4 Argon
• Gemini 3.8 FlashCompare with Gemini 4 Argon