Gemini 4 Argon
A flagship frontier large language model released by Google DeepMind, supporting a maximum output of one million tokens. It is designed for software engineering, legal and financial enterprise knowledge work, and cybersecurity defense scenarios, with capabilities for long-horizon deep reasoning.
Core Facts
Gemini 4 Argon is fully available to paid users, but the Anti-Gravity feature is limited to Ultra subscribers, and Pro users cannot use it.
Gemini 4 Argon is positioned by Google as a 'workhorse' model built specifically for coding and cybersecurity work.
Gemini 4 Argon significantly reduces the hedging expressions, repetition and boilerplate common to such models.
Timeline (last 90 days)
Google's security team's internal benchmarks show that Carbon's success rate in generating secure code is more than 2 times higher than Argon's.
The release of Gemini 4 Carbon comes only two weeks after the release of the previous generation, Gemini 4 Argon.
Compared with Argon, Carbon's unsupported claims are reduced by about 2 times, overly overreaching behavior is reduced by 26%, and under-serving is reduced by 29%.
Carbon retains Argon's multi-layered safety stack, including model training, real-time abuse defense, prompt injection resistance, and chain-of-thought monitoring.
The scoring stage of the Code2Video Benchmark is handled by Gemini 4 Argon, conducting pairwise comparisons across five dimensions: interactivity, requirement compliance, composition, temporal consistency and production technique.
The computational cost and time cost of the 1000-round iteration were not disclosed, which will affect feasibility in actual production environments.
In this case, a 1000-round iteration loop was run, with the model continuously checking the rendering results and optimizing the code.
Google employees said in a tweet that Gemini 4 Argon has "much stronger grasp" of coding and multi-step tool use.
This case comes from Google internal employees, and specific performance metrics and generalization ability have not yet been made public.
Whether Argon's performance in the Blender ecosystem can transfer to other toolchains still requires more third-party verification.
40 more timeline events
All Facts (20)
Gemini 4 Argon is fully available to paid users, but the Anti-Gravity feature is limited to Ultra subscribers, and Pro users cannot use it.
90%VerifiedGemini 4 Argon is positioned by Google as a 'workhorse' model built specifically for coding and cybersecurity work.
90%VerifiedGemini 4 Argon significantly reduces the hedging expressions, repetition and boilerplate common to such models.
85%VerifiedGemini 4 Argon can be invoked through entry points such as the paid Gemini API, Google AI Studio, Antigravity, Gemini Enterprise, AI Mode in the Gemini app and Search, and Android Studio and Firebase AI Logic.
80%VerifiedArtificial Analysis gave Argon an intelligence index score of 53, on par with GPT-6 Astra.
80%VerifiedGemini 4 Argon's maximum output has been increased from 64,000 tokens to 1 million (referring to the output limit, not the context window)
80%VerifiedAI lowers the execution barrier of game development, not the design barrier; core gameplay design, numerical balancing, and level pacing still need to be controlled by humans.
80%VerifiedGoogle's new-generation model Gemini 4 Argon set a new record on the WANDR benchmark.
80%VerifiedBy default, Gemini 4 Argon's output limit for synchronous requests is 65,536 tokens, and exceeding it returns a 400 error; to generate an output close to 1 million tokens, you must set background: true via the Interactions API and raise max_output_tokens, while also requiring store=true because background interactions rely on persistence.
75%VerifiedGemini 4 Argon is not offered in the API free tier, and production-grade calls require binding a paid account.
75%VerifiedOn Terminal-Bench 4.0, Claude Opus 5.5 leads with 66.4%, followed by GPT-6 Astra at 61.8% and Argon at 57.4%.
75%VerifiedGoogle first handed Argon to trusted security experts for early validation through a project called Fairwind
75%VerifiedArgon agents are migrating Google's internal C/C++ codebase to Rust, scaling to over 800,000 lines of the Fuchsia OS Zircon kernel.
75%VerifiedSecurity firm Wiz has used Argon through the Scan for Good program, and in an early demonstration the model discovered a critical vulnerability in a medical software used by hospitals worldwide that was leaking sensitive personal information.
75%VerifiedGemini 4 Argon focuses on three high-value scenarios: software engineering, legal and financial enterprise knowledge work, and cybersecurity defense.
75%VerifiedThe tiered model strategy essentially allocates inference compute dynamically according to task complexity; lightweight models can achieve inference speeds 5–10 times faster than flagship models at one-tenth the cost.
75%VerifiedThe model ID of Gemini 4 Argon is `gemini-4-argon`.
70%VerifiedGemini 4 Argon does not support turning off the thinking feature.
70%VerifiedGemini 4 Argon is a native multimodal model that from the outset was designed to handle text, images, audio and code within a unified architecture, rather than bolting visual capabilities onto a text model.
70%VerifiedThere are two Gemini API Provider options: the Gemini Developer API allows zero-cost quick onboarding and is suitable for testing and lightweight scenarios, while the Agent Platform Gemini API (formerly Vertex AI) is aimed at enterprise-grade production and incurs costs.
70%