DataAIHub
DataAIHubNews · Research · Tools · Learning
GoogleProprietaryReasoningMultimodalVision

Gemini 4 Argon

Google’s Gemini 4 Argon (gemini-4-argon) is a proprietary reasoning model for complex software engineering, scientific problem-solving, and multi-step agent workflows. It accepts text, image, video, audio, and PDF input, returns text, and uses one 1,048,576-token window shared by input, thinking, and output.

Why Gemini 4 Argon matters

Google’s model page positions this id for those flagship workloads. API billing and Gemini app access are different paths. Availability and Pricing on this page keep them separate. The star scores below are DataAIHub qualitative ratings, not official benchmark results.

Vision · Audio input · Tool calling · Thinking · Coding

Last reviewed: 15 October 2026

Official pricing →

When to choose Gemini 4 Argon

Decision guidance for architects—not a feature list.

Best for

  • Complex software engineering
  • Scientific and mathematical problem solving
  • Autonomous multi-step agents
  • Multimodal inputs that should return text

Avoid if

  • You need Argon on the Gemini API free tier, or in AI Studio on the Google AI Pro plan
  • You need image generation, audio generation, or the Live API
  • You want Gemini 3.8 Flash’s lower Standard token price for volume traffic

Strengths

DataAIHub qualitative scores from 1 to 5, for architects. Not official benchmark results or a public ranking.

  • Hard reasoning / agents★★★★★
  • Long context★★★★★
  • Multimodal input★★★★★
  • API price vs Gemini 3.8 Flash★★☆☆☆
  • Open weights★☆☆☆☆

Ecosystem

Built by

  • Google

    Gemini 4 Argon is Google’s public flagship model as of 2026-10-13.

Competes with

  • GPT-5.6

    Competes on frontier coding, reasoning, and multimodal APIs.

  • Claude Opus

    Competes for long-running agents and hard professional work.

Works with

  • Gemini API

    Paid Gemini API id gemini-4-argon in AI Studio and Android Studio.

  • GitHub Copilot

    Generally available in Copilot for Pro+, Max, Business, and Enterprise from 2026-10-13.

Recommended for

  • Gemini Models

    How Argon, Flash, and Pro-class IDs differ when you pin a Gemini model.

  • AI Agents

    Built for autonomous multi-step agents. Long decode can extend output only within the shared 1,048,576-token window.

Often paired with

  • Gemini 3.1 Pro

    3.1 Pro remains the scored Pro-class pin. Argon is a newer public id, not that slug.

  • Gemini Flash

    Gemini 3.8 Flash stays the lower-priced workhorse under Argon.

How Gemini 4 Argon evolved

Key moments in chronological order.

  1. Model

    Gemini 4 Argon

    Google releases Gemini 4 Argon (gemini-4-argon). Paid Gemini API access uses Cloud Billing and is not on the API free tier. A Google AI Pro subscription does not grant that API access. In AI Studio, Argon is not on the Google AI Pro plan. Google AI Pro and Ultra subscribers can still use it in the Gemini app and AI Mode in Search. The 1,048,576-token window is shared by input, thinking, and output; it is not a separate 1M-token output allowance. Standard API price is $2 input / $10 output per 1M tokens through 2026-11-30, then $4 / $20 from 2026-12-01. Fairwind is a restricted program for Argon without cyber guardrails, not a second public model id.

Overview

Stable Gemini API id gemini-4-argon, released 13 October 2026. Google’s model page describes it as the flagship for complex software engineering, deep mathematical and scientific problems, autonomous multi-step agents, and long decode. Thinking can be set to low, medium, or high. The long-decode guide names medium as the default for most coding and agentic work. It documents function calling, structured outputs, code execution, search grounding, URL context, and file search. Computer use is in preview. The model page does not document native MCP support. Image generation, audio generation, and the Live API are not supported.

Capabilities

  • Vision: Yes
  • Audio input: Yes
  • Tool calling: Yes
  • Thinking: Yes
  • MCP: Not documented
  • Coding: Yes
  • Structured output: Yes

Technical specifications

Provider
Google
License
Proprietary
Context window
Shared by input, thinking, and output
Parameters
Undisclosed
Architecture
Undisclosed (proprietary)
Release
2026-10
Modalities
Text, Image, Audio, Video, PDF
Vision
Yes
Audio input
Yes
Tool calling
Yes
Thinking
Yes
MCP
Not documented
Open weights
No
API
Yes
Pricing (input)
$2 per 1M tokens through 2026-11-30; $4 from 2026-12-01 (gemini-4-argon Standard)
Pricing (output)
$10 per 1M tokens through 2026-11-30; $20 from 2026-12-01 (thinking tokens included)

Supported modalities

Inputs: Text · Image · Audio · Video · PDF

Output: Text

Context window

Shared by input, thinking, and output (1,048,576 tokens)

Pricing

Input: $2 per 1M tokens through 2026-11-30; $4 from 2026-12-01 (gemini-4-argon Standard)
Output: $10 per 1M tokens through 2026-11-30; $20 from 2026-12-01 (thinking tokens included)

These are Gemini API rates, not Gemini app subscription prices. Output price includes thinking tokens. Cached input is implicit only: $0.10 per 1M through 2026-11-30, then $0.20. Explicit caching is not supported. Batch and Flex: $1.00 input / $5.00 output through 2026-11-30, then $2.00 / $10.00; cached input $0.05, then $0.10. Priority: $3.60 / $18.00 through 2026-11-30, then $7.20 / $36.00; cached input $0.18, then $0.36. The pricing page says Argon is not on the API free tier or the Google AI Pro plan. That plan is an AI Studio subscription, separate from pay-as-you-go API billing, and that note does not describe Gemini app access. On paid Standard, Batch, Flex, and Priority, “Used to improve our products” is No. The free-tier column is Not available on every Argon row. Contributor is an invite-only private preview, not a public rate: the logs policy says enabling it permits Google to use that project’s participating-model data to provide, improve, and develop Google products and machine-learning technologies, under the Unpaid Services terms, in exchange for up to 50% off. The pricing table marks that paid tier Yes, at the same $1.00 / $5.00 then $2.00 / $10.00 token rates as Batch and Flex.

Availability

API: Yes
Chat UI: Yes
Open weights: No

Gemini API

Pay-as-you-go Gemini API access, including an API key in AI Studio and Android Studio, uses Cloud Billing. Argon is not on the API free tier. A Google AI Pro or Ultra subscription does not grant that API access. The plans page says direct API use is billed separately. Enterprises can use Gemini for business, the API, and Antigravity.

Gemini app and Google AI subscriptions

Google AI Pro and Ultra are AI Studio subscription plans, separate from API billing. In AI Studio, Argon is not on the Free or Google AI Pro plan, and Google AI Ultra includes it. Those plan benefits apply in the AI Studio web UI. The 13 October 2026 launch post says Google AI Pro and Ultra subscribers can use Argon in the Gemini app and AI Mode in Search, and Ultra subscribers can try it in Antigravity. The plans page says the Antigravity Preview agent inside AI Studio is not part of Google AI plans and needs a paid API key.

IDE and third-party integrations

GitHub Copilot lists it for Copilot Pro+, Max, Business, and Enterprise, with a gradual rollout, billed at provider list price. Business and Enterprise admins can disable it in model policy.

Restricted-access programs

Fairwind is an application program for Argon without cyber guardrails, on a rolling basis. It is not a second public model id. The same launch post says Google is offering up to 10,000 scientists the highest Antigravity usage tier at no charge for 6 months, onboarded on a rolling basis and subject to eligibility, capacity, and selection. An application does not guarantee a place.

Long-decode limits

Without long decode, streaming calls and current Google GenAI SDK calls stop at 64K output tokens, and non-streaming REST calls stop at 32K. With long decode, output can use only the shared window that remains after input. The model page also lists a 65,536-token default synchronous max output. One hop returns at most 65,536 tokens (32,768 on non-streaming REST).

Use cases

  • Complex software engineering and multi-step agents
  • Scientific and mathematical problem solving
  • Documents, images, video, and audio with text output
  • Paid Gemini API calls, or the Gemini app on Google AI Pro and Ultra

Strengths

  • Stable API id gemini-4-argon, released 13 October 2026
  • 1,048,576-token window shared by input, thinking, and output
  • Documented function calling, structured outputs, and code execution

Limitations

  • Closed weights
  • Not on the Gemini API free tier. In AI Studio, not included on the Free or Google AI Pro plan. A Google AI Pro subscription can still reach Argon in the Gemini app and AI Mode in Search
  • Standard API list price is higher than Gemini 3.8 Flash introductory rates, and Argon Standard rates rise on 2026-12-01
  • No image generation, audio generation, or Live API on this id
  • Explicit context caching is not supported
  • The Gemini 4 Argon model page does not document native MCP support

Related guides

Related benchmarks

Related research

Related GitHub

Related tools

Related rankings

Related comparisons

Companies

Explore more models

All models →