Is Gemini 4 Argon available with Zero Data Retention on Elosia?

Gemini 4 Argon runs on Elosia's standard backend; ZDR is available on many other models in the catalog.

How much does Gemini 4 Argon cost on Elosia?

Gemini 4 Argon is billed by usage at $2 per 1M input tokens and $10 per 1M output tokens, within your Elosia plan.

What is Gemini 4 Argon's context window?

Gemini 4 Argon supports a context window of up to 1,048,576 tokens.

Is Gemini 4 Argon hosted in the EU?

Yes, Gemini 4 Argon is available on Elosia's EU-hosted (EdenAI) backend for enterprise organizations.

When should you choose Gemini 4 Argon?

Gemini 4 Argon is a flagship model, best for the most demanding reasoning, agentic and long-context tasks.

Does Gemini 4 Argon support image inputs on Elosia?

Yes, Gemini 4 Argon accepts image inputs; image tokens are priced separately (see the pricing table).

Gemini 4 Argon

GeminiFlagship

ThinkingTool UseVisionStructured Output

About this model

Gemini 4 Argon is Google's flagship reasoning model, built for complex software engineering, deep mathematical and scientific problem solving, and autonomous multi-step agents. Google trained it to keep context across...

Performance Tier

Flagship

Gemini 4 Argon is a flagship model from Gemini : the most capable in their lineup.

Best-in-class model from this provider. Highest performance across benchmarks, ideal for demanding tasks.

Pricing

This model is included in Elosia plans

High

High pricing. Reserve this level for tasks that demand maximum quality.

Type	per 1M tokens
Input (prompt)	$2.00
Output (completion)	$10.00
Image	$2.00
Internal reasoning	$10.00
Cache read	$0.100
Cache write	$0.188

Capabilities

Context Length1.0M

Max Output Tokens66K

TokenizerGemini

Inputaudio, file, image, text, video

Outputtext

Release DateOctober 13, 2026

Benchmarks

General Intelligence

MMLU

Not reported

GPQA Diamond

Not reported

Mathematics

MATH-500

Not reported

Programming

HumanEval

Not reported

SWE-bench Verified

Not reported

Reasoning

IFEval

Not reported

Humanity's Last Exam

57.1%

Agentic

Terminal-Bench 4.0

57.4%

Where does Gemini 4 Argon stand?

Compare its performance index against every other model.

View the performance leaderboard

Recommended Use Cases

CodingResearchAnalysisMathematicsData ExtractionSummarization

Strengths

Google's flagship reasoning model, built for complex software engineering, maths and science, and autonomous multi-step agents
Vals Index 68.9%, ahead of GPT-6 Astra, Claude Fable 5.1 and Claude Opus 5.5 in Google's reported comparison
Leads Finance Agent v2 (65.4%), Vibe Code Bench (91.9%) and DeepSWE v1.1 (77.9%) among the compared frontier models
1M-token unified context window with 64K output on Elosia
Natively multimodal input covering text, images, video, audio and PDFs, with strong long-video understanding (LVBench 91.7%)
15.1% hallucination rate on AA-Omniscience, far below GPT-6 Astra (51.3%) and Claude Opus 5.5 (58.6%) as measured by Artificial Analysis

Limitations

Thinking has no off setting (low, medium or high only) and produces long reasoning outputs, which raises latency and output cost
Explicit context caching is not supported, only implicit caching
Google publishes no GPQA Diamond, MMLU, AIME or SWE-bench Verified score; its reported suite is built around agentic and domain evaluations
Google's figures are self-reported and the Humanity's Last Exam score comes from Artificial Analysis; Argon ranks last of the four models Google compares on Terminal-Bench 4.0 and on FrontierSWE v2
AA-Omniscience factual accuracy of 50%, below Gemini 3.1 Pro Preview and GPT-6 Astra