stealthmodels.

MODEL BENCHMARKS

Gemini 4 Argon: Ground Truth

Google’s multimodal reasoning model combines sparse mixture-of-experts architecture with sustained reasoning and a shared million-token input/output context.

A good fit for

Complex software engineering, scientific and mathematical problem solving, and long-running agents working across documents, images, audio and video.

Context window
1,048,576tokens[5] [6]
License
Proprietary hosted service[4] [1]
Model factsArchitecture, context & capabilities
First available
2026-10-13[1] [2]
Developer
Google[1] [2]
Origin
United States[3]
Family
Gemini 4[1] [2]
Context window
1,048,576 tokens[5] [6]
Maximum output
943,718 tokens[6]
Architecture
Sparse mixture-of-experts (MoE) transformer with native multimodal inputs[1]
Input
text, image, audio, video, PDF[5]
Output
text[5]
Normal speed
143.66 tokens/sec avgMin 138.69 · Max 147.61 tokens/secOur measurement
API access & pricingOpenRouter · $2 input / $10 output · per 1M tokens

OpenRouter / Gemini API Standard: $2 input, $0.10 cached input and $10 output per 1M tokens through 2026-11-30; Gemini API Standard becomes $4 / $0.20 / $20 on 2026-12-01. Output includes thinking tokens. Flex is a separate tier at $1 input / $0.05 cached input / $5 output during the introductory period.

[8] [6]
  • OpenRouterObserved 2026-10-13
    Input $2Output $10Cached read $0.1

    per 1M tokens

    google/gemini-4-argon[7]
  • Gemini Developer APIObserved 2026-10-13
    Input $2Output $10Cached read $0.1

    per 1M tokens

    Available

    Standard introductory price through 30 November 2026; output includes thinking; implicit caching only.

    gemini-4-argon[8]
  • Gemini Developer APIObserved 2026-10-13
    Input $4Output $20Cached read $0.2

    per 1M tokens

    announced-future-rate

    Announced Standard price effective 1 December 2026.

    gemini-4-argon[8]
  • Gemini Developer API / FlexObserved 2026-10-13
    Input $1Output $5Cached read $0.05

    per 1M tokens

    Available

    Flex introductory price through 30 November 2026; separate service tier.

    gemini-4-argon[8]
Measurements & sources8 primary references

StealthMark score

StealthMark measures fluid and visual intelligence of AI models, resisting any pre-existing benchmark focused training.

General and Visual

General Intelligence

The ability to reason, solve problems and understand how things relate and behave.

Visual Intelligence

The ability to imagine, create and accurately reconstruct complex images in code form.

Judging

Scores reflect how well each result meets the benchmark requirements.

Overall score

Overall combines weighted task scores on a scale from 0 to 100. Estimated means the total includes an estimate for one missing scene.

Scoring categories

Task & likeness

Subjects, actions and likeness.

Anatomy & construction

Coherent bodies, joints and machinery.

Space & placement

Perspective, scale and contact.

Light & reflections

Lighting, shadows and reflections.

Purposeful detail

Clear, purposeful small features.

Artistry & character

Composition, expression and drawing skill, including complexity and its execution.

Materials & effects

Surfaces, texture and movement.

Visual integrity

Clean shapes, layers and edges.

BENCHMARK RESULTS

Compare results

This link opens the same results in the same order.

StealthMark benchmark results

Reset benchmark preferences?

Return to the default view, clear comparison selections and restore the advertisement.