Gemini 4 Argon
Gemini 4 Argon is Google's frontier model designed for long-horizon coding, legal and finance work, and cyber defense, with a 1M token output limit.
Model Overview
Capabilities, design details, and architectural traits
Gemini 4 Argon - Google's next era of frontier intelligence
Gemini 4 Argon is Google's frontier model built to sustain deep reasoning across complex, long-horizon workflows. It targets real-world software engineering, enterprise knowledge work such as legal and finance, and cybersecurity defense. It also introduces a new naming scheme for Google's model lineup.
| Defining trait | What it means |
|---|---|
| 1M token output limit | Expanded from the previous 64K, so the model can think deeply and generate hundreds of thousands of tokens in a single trajectory to solve tough problems in one go. |
| Cybersecurity defense focus | Trained to autonomously find, validate, and patch critical software vulnerabilities, and made available without cyber guardrails to trusted defenders and internal Google teams. |
| Fairwind Program rollout | First released to a limited set of trusted cyber defenders through Google's Fairwind Program, with more than 650 participating partners globally. |
| Misalignment monitoring | Deployed mitigations monitor Argon's chain-of-thought and actions and stop execution when necessary, with precautions against feeding monitoring findings back into training. |
Changing how Google works and builds
Argon is already powering internal workflows for thousands of Googlers. Teams of Argon agents analyzed fleet-wide profiling telemetry to autonomously apply memory optimizations across Google's data centers, freeing over 300 TiB of memory. Other Argon agents are migrating C/C++ codebases to Rust, scaling from core libraries like re2 and libgav1 up to 800K+ lines of the Fuchsia Zircon kernel, with all rewrites undergoing rigorous auditing and review before production.
Phased release under frontier safeguards
Google is engaged in the U.S. government's voluntary process for pre-release model access while gradually expanding availability. Before broader release, the company is hardening sandboxed environments, monitoring internal activations to spot misuse, and strengthening resilience against indirect prompt injection attacks. Wider access will start with Google AI Ultra subscribers and paid API customers.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Anthropic | Anthropic | |
| Release Date | September 30, 2026 | September 22, 2026 | September 28, 2026 |
| Knowledge Cutoff | - | - | Jun 2026 |
| Context & Limits | |||
| Context Window | - | 1M | 1M |
| Pricing (per 1M tokens) | |||
| Input Pricing | $2 Best Input Pricing | $4 | $2 Best Input Pricing |
| Output Pricing | $10 Best Output Pricing | $20 | $10 Best Output Pricing |
| Modalities | |||
| Inputs | textimagefilevideo | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 52.6 | 57.6 Best Intelligence Index | 56.0 |
| Coding Index | - | - | - |
| Agentic Index | - | - | - |
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
SciCode Benchmark
Scientific coding and mathematical modeling.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Explore more from Google
Other models by Google
Gemini 3.8 Flash
Gemini 3.7 Flash
Gemini 3.6 Flash
Gemini 3.5 Flash
Top AI Models
Leading alternatives by intelligence score
Claude Opus 5.5
Muse Spark 1.3
GPT-6 Astra
