Google Unveils Gemini 4 Argon, its Most Powerful AI Model Yet

Google has unveiled Gemini 4 Argon, its most capable AI model, topping 13 of 18 benchmarks while targeting software engineering, legal, finance and cybersecurity tasks with expanded output limits and new safeguards.
Google unveils Gemini 4 Argon, its most powerful AI model yet; CEO Sundar Pichai says_ We're going to make it available as soon as we can and as
Written By:
Somatirtha
Reviewed By:
Achu Krishnan
Published on: 
Updated on: 

Google has unveiled Gemini 4 Argon, its most capable artificial intelligence model so far and the first model in the Gemini 4 series. Announced on September 30, Argon is designed to handle long, multi-step tasks across software engineering, legal and finance work, and cybersecurity defense.

The launch follows Google’s decision to quietly drop Gemini 3.5 Pro, which the company had promised for June.

Gemini 4 Argon Tops 13 of 18 Benchmarks

Argon currently tops 13 of 18 benchmarks against models from OpenAI and Anthropic. On DeepSWE v1.1, which tests long-horizon software engineering, Argon scored 77.9%, compared with 74.2% for Claude Opus 5.5 and 74.1% for GPT-6 Astra.

On Harvey’s Legal Agent Benchmark, Argon scored 19.6%, ahead of Astra’s 5.4%. It also recorded 68.9% on the Vals Index, which measures finance, coding, legal, and tax work based on their respective shares of U.S. GDP. 

However, GPT-6 Astra leads Argon on FrontierSWE v2 and OSWorld-2.0, while Claude Opus 5.5 leads on Terminal-bench 4.0 and PostTrainBench. Argon and Astra are tied at 68% on CWE-bench v1.

Google has also increased Argon’s output limit from 64,000 to 1 million tokens. Introductory API pricing is USD 2 per million input tokens and USD 10 per million output tokens, while cached input is 95% cheaper.

Google Already Using Argon Internally

Thousands of Googlers have been using Argon internally for weeks. A team of Argon agents freed more than 300 TiB of memory across Google’s data centers, with total savings estimated at 500 TiB to 1 PiB.

The agents are also moving C/C++ codebases to Rust, including more than 800,000 lines of the Fuchsia Zircon kernel. In the libgav1 video decoder, Argon replaced 32,000 lines of SIMD code, making the result 2.7 times faster than the earlier Rust port.

Also Read: ChatGPT vs Gemini: How Their Visual AI Capabilities Compare

Wider Release to Follow

Argon is initially being provided to trusted cyber defenders through Google’s Fairwind Program while it goes through the U.S. government’s voluntary pre-release access process. Paid API customers and Google AI Ultra subscribers will follow, although Google has not announced a date.

CEO Sundar Pichai said, “We’re going to make it available as soon as we can and as safely as we can.”

Google has also added safeguards against cyberattacks, CBRN weapons misuse, and indirect prompt injection before a wider release.

Join our WhatsApp Channel to get the latest news, exclusives and videos on WhatsApp
logo
Artificial Intelligence News & Cryptocurrency News: Latest Trends | Analytics Insight
www.analyticsinsight.net