Google unveils Gemini 4 Argon, claims top AI benchmarks

October 1, 2026 at 01:11 AM2 min

This digest was compiled by AI from multiple sources — links to the originals are below.

Google announced Gemini 4 Argon on September 30, its most advanced AI model, claiming top scores in coding, cybersecurity, and professional work benchmarks. The model is already used internally to save 300 TiB of memory and migrate over 800,000 lines of code to Rust. Google will roll it out first to trusted cybersecurity partners, with no public release date or pricing yet.

Key Facts

Gemini 4 Argon scores 77.9% on the DeepSWE v1.1 software engineering benchmark, ahead of GPT-6 Astra (74.1%) and Claude Opus 5.5 (74.2%).
The model supports a 1 million token output limit, up from 64,000 tokens in previous Gemini models.
Google used Argon to save 300 TiB of memory across its data centers and migrate over 800,000 lines of Fuchsia OS Zircon kernel code to Rust.
Google will roll out Gemini 4 Argon in phases, starting with trusted cybersecurity partners and working with the U.S. government on pre-release safety evaluations.
CEO Sundar Pichai signed a voluntary AI safety accord with President Donald Trump and major tech leaders this week.

Benchmark Performance

Google claims Gemini 4 Argon sets a new record in real-world software engineering, ties for first in cybersecurity, and leads a benchmark measuring performance across finance, legal, and other professional tasks. On the DeepSWE v1.1 benchmark, Argon scores 77.9%, compared with Claude Opus 5.5 at 74.2% and GPT-6 Astra at 74.1%. Google also points to Argon's industry-leading score in the economic analysis Vals Index test. The model is described as built for deep reasoning across complex, long-horizon workflows, with expanded coding, reasoning, and multimodal capabilities.

Internal Deployment

Google engineers are already using Gemini 4 Argon extensively inside the company. Argon used fleet-wide telemetry data to help Google save 300 TiB of memory across its data centers. Argon agents migrated C/C++ codebases to Rust, including thousands of lines in the core re2 and libgav1 libraries and more than 800,000 lines in the Fuchsia OS Zircon kernel. Quantum computing researchers at Google have also utilized the model.

Rollout and Safety

Google plans to launch Gemini 4 Argon in phases, starting with trusted cybersecurity partners. The company is working with the U.S. government on pre-release safety evaluations. Tulsee Doshi, Google's Gemini model product lead, said starting the rollout with cyber defenders gives Google more confidence and puts a model trained in cyber defense in the hands of defenders as soon as possible. The model will be made available without cyber guardrails to trusted defenders and internal Google teams to leverage its full frontier-level cybersecurity defense capabilities. Google has not announced API pricing or a public release date for Gemini 4 Argon.