Claude Opus 5.5 is live on SeedRouter
Google DeepMindFrontier model

Gemini 4 Argon — Google's frontier model for coding and cyber defense

Gemini 4 Argon is Google’s frontier model for long-horizon coding, document analysis and multimodal reasoning. Use native Gemini requests with text, images, audio, video and PDFs, and receive text responses. $0.90 / 1M input · $4.50 / 1M output.

Available$0.90 / 1M input · $4.50 / 1M output$2.00 / 1M input · $10.00 / 1M output-55%
  • 1M-token shared context
  • Long-horizon coding
  • Structured output
  • Function calling
  • Multimodal input
Developer
Google DeepMind
Announced
September 30, 2026
Shared context
1,048,576 tokens
API model ID
gemini-4-argon

Playground

Use native Gemini requests with your account balance. Basic text streaming has been verified locally; individual parameter behavior is still being checked.

Parameters
gemini-4-argon$0.90 in · $4.50 out per 1M tokens
thinkingLevel
Conversation total$0.000000
Ready

Try one of these prompts

Messages are billed from your balance at the rates below, like an API call. Failed requests are free.

Gemini 4 Argon pricing

Gemini 4 Argon rates are read from current billing configuration. Input, cached input and output are priced separately; output includes thinking tokens.

One typical request2,000 input tokens, 1,000 output tokens
$0.0063per request
Thinking counts as output, so higher effort raises the output side.
What $20.00 buysOutput tokens at the current rate
4,444,444output tokens
Input and cache tokens cost less per token than output.

Example sizes are for illustration. Your bill follows the tokens each request reports. Cached tokens are part of input, not additional input. Each continuation request has its own usage. Failed requests are not charged.

CategoryDefaultgemini-4-argongoogle llm vertextgemini-4-argon:googlellm-vertext
Price · 1M tokensPrice · 1M tokens
Input$0.90$1.40
Output, including thinking$4.50$7.00
Cached input$0.045$0.07

What is Gemini 4 Argon?

Gemini 4 Argon is Google's new frontier model, announced on September 30, 2026. Google built it to sustain deep reasoning across complex, long-horizon workflows, and positions it for real-world software engineering, enterprise knowledge work such as legal and finance, and cybersecurity defense.

Gemini 4 Argon shares a 1,048,576-token window between input, thinking and output. Native long responses can span several continuation requests. Google announced wider availability on October 13, 2026; the API model ID is gemini-4-argon.

Gemini 4 Argon announced September 30, 2026 · Wider release October 13, 2026 · Native Gemini API

REST output per request
32,768 tokens; 65,536 streaming
Context window
1,048,576 shared tokens
Multimodal
Text, image, audio, video and PDF input
API model ID
gemini-4-argon
Output
Text only

Two ways to use Gemini 4 Argon

Use Gemini 4 Argon from your application or coding agent.

API

Call it from your backend

Best for apps and services

Use the native Gemini request format for Gemini 4 Argon. The model ID belongs in the endpoint path.

  1. 1Create an API key
  2. 2Put gemini-4-argon in the endpoint path
  3. 3Send messages in contents and settings in generationConfig
Agent

Hand it to your coding agent

Best for Claude Code, Codex and other agents

Copy the prompt, then review the request and estimated cost in your agent.

  1. 1Copy the prompt
  2. 2Paste it into your agent
  3. 3Approve the request

What Gemini 4 Argon is built for

The Gemini 4 Argon strengths Google highlights in its announcement, with the scores it published.

Long-horizon coding

Google reports 77.9% on DeepSWE v1.1, a benchmark of real-world, long-horizon software engineering tasks, and uses Gemini 4 Argon agents for large codebase migrations.

Enterprise knowledge work

Gemini 4 Argon leads the Vals Index at 68.9%, which weights finance, coding, legal and tax work by contribution to U.S. GDP, and scores 51.3% on Zapier's AutomationBench.

Cybersecurity defense

Google trained Gemini 4 Argon to find, validate and patch software vulnerabilities on its own. It ties for first place on CWE-bench v1 with 68%.

Long responses with continuation

Gemini 4 Argon can use the remaining shared context for thinking and output. Longer native generations return a continuation token; each continuation is another request.

Charts, documents and video

Google reports a 91.7% Gemini 4 Argon score on LVBench for long-video understanding, and highlights chart analysis and work across a series of documents.

Prompt-injection resistance

Google calls Gemini 4 Argon its most resilient model yet against indirect prompt injection, and says it leads Gray Swan's Indirect Prompt Injection benchmark.

Gemini 4 Argon use cases Google has described

Gemini 4 Argon examples from Google's own teams and early partners, as reported by Google.

Codebase migration

Gemini 4 Argon agents are migrating C and C++ code to Rust across Google, from core libraries such as re2 up to the 800K+ lines of the Fuchsia Zircon kernel.

Performance engineering

In libgav1, Gemini 4 Argon agents replaced 32K lines of SIMD code with safe Rust. The decoder runs 2.7x faster than the earlier Rust port with identical video output.

Infrastructure optimization

A team of Gemini 4 Argon agents found and applied memory optimizations across Google's data centers, freeing over 300 TiB of memory once rolled out.

Vulnerability research

Wiz used Gemini 4 Argon to uncover a critical vulnerability in healthcare software used by hospitals worldwide, a risk earlier frontier models had missed.

Finance and legal research

In Google's results, Gemini 4 Argon leads Vals Finance Agent v2 with 65.4% and Harvey's Legal Agent Benchmark with 19.6%.

Algorithm design

Google's quantum researchers used Gemini 4 Argon to optimize a subroutine's qubit and gate cost, beating the published baseline by 40% in minutes.

How to use Gemini 4 Argon

  1. 01

    Create an API key

    Sign in to SeedRouter, add credit and create an API key for Gemini 4 Argon. Keep the key on your server.

  2. 02

    Choose a native operation

    Call Gemini 4 Argon with generateContent or streamGenerateContent, using gemini-4-argon in the URL. Put the prompt and conversation in contents.

  3. 03

    Set your generation options

    Choose a Gemini 4 Argon thinking level and an output budget. Add media, a response schema or tools when your task needs them.

  4. 04

    Inspect the result

    Read Gemini 4 Argon candidates, finishReason and usageMetadata. Keep thought signatures and tool parts unchanged in the next conversation turn.

Try in Playground

Gemini 4 Argon API

The Gemini 4 Argon reference covers native requests, nested parameters, streaming and continuation. Capability verification is still in progress.

Native model ID

Use gemini-4-argon in the model path. The Gemini 4 Argon JSON body uses contents and generationConfig rather than a custom prompt format.

Complete native parameters

Configure Gemini 4 Argon thinking, structured output, tools, multimodal parts, safety settings, labels and continuation. Officially ignored sampling settings are not active controls.

Usage-based billing

Gemini 4 Argon input includes cached tokens; output includes thinking. Each continuation request reports its own usage. Failed requests are not charged.

Gemini 4 Argon benchmarks

Google's published results against GPT-6 Astra, Claude Fable 5.1 and Claude Opus 5.5. Across the 19 rows shown below, Gemini 4 Argon leads on 13, ties on CWE-bench v1 and trails on 5.

BenchmarkGemini 4 ArgonGPT-6 AstraClaude Fable 5.1Claude Opus 5.5
Vals Index · knowledge work68.9%63.1%65.8%67.0%
AutomationBench · knowledge work51.3%41.4%31.4%42.5%
Vals Finance Agent v2 · knowledge work65.4%53.5%58.9%58.6%
Harvey's Legal Agent Benchmark · knowledge work19.6%5.4%6.7%3.8%
DeepSWE v1.1 · agentic coding77.9%74.1%67.4%74.2%
FrontierSWE v2 · agentic coding55.0%65.5%56.3%62.3%
Vibe Code Bench · agentic coding91.9%89.6%90.3%90.3%
Terminal-bench 4.0 · agentic coding57.4%58.2%57.9%66.4%
PostTrainBench · ML engineering45.3%44.3%40.2%49.3%
Terminal-Bench Science 0.1 · science57.6%68.1%52.6%63.3%
LABBench 2 · science88.8%85.4%68.6%73.1%
RiemannBench · math76.0%72.0%65.6%69.6%
GraphWalks up to 128K · long context99.7%98.7%91.4%90.6%
GraphWalks 256K to 1M · long context84.2%71.8%65.0%66.8%
Agent's Last Exam · computer use39.5%34.2%—38.2%
OSWorld-2.0 offline subset · computer use69.2%72.6%——
Chartography · multimodal71.6%71.0%46.2%66.3%
LVBench · long video91.7%87.5%79.7%83.7%
CWE-bench v1 · cybersecurity68.0%68.0%58.0%67.0%

Scores as published by Google on September 30, 2026, with Gemini 4 Argon at its highest thinking setting. Results for other models are mostly their developers' own reported numbers; SeedRouter has not run these benchmarks. Source: Google DeepMind, Gemini 4 Argon

Gemini 4 Argon limits and verification

Some capabilities still need verification

Basic Gemini 4 Argon text streaming has passed a local end-to-end check. Tools, media, continuation and individual parameter behavior still need verification.

One shared context window

Gemini 4 Argon input, thinking and output share 1,048,576 tokens. A large output budget does not provide an additional input window.

No image or audio generation

Gemini 4 Argon accepts multimodal input but produces text. Live audio and explicit cachedContent resources are not supported by this model.

Benchmarks are Google's

The table above is Google's own. Gemini 4 Argon trails GPT-6 Astra or Claude Opus 5.5 on FrontierSWE v2, Terminal-bench 4.0, PostTrainBench, Terminal-Bench Science and OSWorld-2.0.

Preserve safety and conversation state

Keep Gemini 4 Argon thought signatures and tool parts intact. Safety settings and tool permissions should match your application’s needs.

Why use Gemini 4 Argon on SeedRouter

One key for many models

Use one SeedRouter API key for Gemini 4 Argon, GPT-6, Claude, Kimi and DeepSeek text models, plus image and video models.

Pay per use

Spend your SeedRouter balance on Gemini 4 Argon or other listed models. No plan, no monthly fee.

Failures are free

Failed Gemini 4 Argon requests are not charged.

Live prices

Gemini 4 Argon rates on this page come from current billing settings.

Models you can call today

Compare Gemini 4 Argon with other text models available on SeedRouter at their current rates.

GPT-6 Astra
gpt-6-astra

Call GPT-6 Astra with the OpenAI SDK. OpenAI's most capable model, with a 1.05M-token context, live per-token pricing and a playground you can try right now.

View pricing
GPT-6 Sol
gpt-6-sol

Call GPT-6 Sol with the OpenAI SDK. 1.05M-token context, reasoning effort from none to max, live per-token pricing and a playground you can try right now.

View pricing
GPT-6.1 Sol
gpt-6.1-sol

Call GPT-6.1 Sol with the OpenAI SDK: near-Astra coding and agent work at a fifth of the price, 1.05M-token context, 95% off cached input. Try it now.

View pricing
GPT-6 Luna
gpt-6-luna

Call GPT-6 Luna with the OpenAI SDK. The lowest-cost GPT-6 model, with a 1.05M-token context, live per-token pricing and a playground you can try right now.

View pricing
Claude Fable 5.2
claude-fable-5-2

Try Claude Fable 5.2 for coding and problem solving with a 1M-token context. Compare live token prices and model tiers, then test it in the playground.

View pricing
Claude Mythos 5
claude-mythos-5

Try Claude Mythos 5 for complex reasoning and research with a 1M-token context. Compare live token prices and model tiers, then test it in the playground.

View pricing

Gemini 4 Argon guides

Read Gemini 4 Argon release coverage, integration guides and comparisons.

Gemini 4 Hydrogen vs Argon: Which Model Should You Use?

Compare Gemini 4 Hydrogen and Argon for coding, agents and long documents: thinking defaults, output limits, current prices and three measured API checks.

Gemini 4 Carbon vs Argon: which Gemini 4 model should you use?

Gemini 4 Carbon and Argon share one request format and price. We ran the same three prompts on both through the API: thinking tokens, latency and answers.

Gemini 4 Argon 3D code: reproduce the official Python scene

Run the official Gemini 4 Argon 3D code on a Mac with Python and NumPy. Reproduce a sunlit library, inspect the result and change rendering parameters.

Gemini 4 Argon API: a Python chat with native streaming

Build a Gemini 4 Argon API conversation in Python. Read native SSE, preserve signed content parts, handle empty answers and inspect token usage.

Gemini 4 Argon in OpenCode: configuration and current limits

Configure Gemini 4 Argon with OpenCode’s native Google provider. See the tested version, signed-history behavior and unresolved continuation limits.

Gemini 4 Argon pricing: calculate the cost of a request

Estimate Gemini 4 Argon costs from input, cached input and thinking tokens. Read current SeedRouter rates and budget for conversations and long responses.

Gemini 4 Argon vs Claude Opus 5.5: which fits your task?

Compare Gemini 4 Argon and Claude Opus 5.5 using official context limits, thinking controls, benchmarks and native API differences before choosing a model.

Gemini 4 Argon Release Date: Is It Out, and Who Has Access?

Read the current Gemini 4 Argon access update for Gemini App, Antigravity, AI Studio and API users, alongside the dated original announcement.

Gemini 4 Argon FAQ

Is Gemini 4 Argon out?

Yes. Google announced Gemini 4 Argon on September 30, 2026 and wider availability on October 13, 2026. The official API model ID is gemini-4-argon.

When will Gemini 4 Argon be available?

Google announced wider availability on October 13, 2026. Check Google’s current availability information for your account and region.

How much does Gemini 4 Argon cost?

Use the live pricing section on this page for SeedRouter’s current input, cached-input and output rates. Output billing includes thinking tokens. Each continuation request is billed separately.

What is Gemini 4 Argon?

Gemini 4 Argon is Google’s frontier model for complex coding, knowledge work and multimodal reasoning. Its 1,048,576-token context is shared by input, thinking and output.

How does Gemini 4 Argon compare with GPT-6 Astra and Claude Opus 5.5?

Across the 19 rows in the comparison table on this page, Gemini 4 Argon leads on 13, ties GPT-6 Astra on CWE-bench v1 and trails on 5, including FrontierSWE v2 and Terminal-bench 4.0. These are Google's own results.

Is there a Gemini 4 Argon API?

Yes. Gemini 4 Argon uses the native Gemini generateContent and streamGenerateContent operations. The linked API reference lists fields, constraints, examples and current verification limits.

Can I use Gemini 4 Argon on SeedRouter?

Gemini 4 Argon is configured and visible through the API. This page includes its native Playground and API documentation. Basic text streaming has been verified locally; individual parameter checks remain in progress.

What is Gemini 4 Argon's output limit?

Native REST responses have a per-request ceiling of 32,768 tokens, or 65,536 when streaming. Longer generations use continuation tokens within the shared 1,048,576-token context. Thinking consumes the output budget too.

What is Google's Fairwind Program?

Fairwind is Google DeepMind’s program for approved cyber defense organizations. It provided early Gemini 4 Argon access before the wider October 13 release.

Who makes Gemini 4 Argon?

Gemini 4 Argon is made by Google. Google DeepMind develops the Gemini models.

Build with Gemini 4 Argon

Explore the native Gemini 4 Argon request format and read the current capability notes before integrating.