Google's Gemini 4 Argon targets the hardest problems first
The new frontier model from Google DeepMind is already helping engineers and scientists tackle months of work in weeks.

A program that reads huge documents and writes code so scientists and engineers finish months of work in weeks.
- Connect to Google's platform or app
- Paste your documents or code
- Tell it what problem to solve
- Get results in days instead of months
A new model for deep work
Google DeepMind has launched Gemini 4 Argon, a new artificial intelligence model built for complex, real-world tasks. It is now available to developers, enterprises, scientists, and consumers through Google's platforms. The model carries an industry-leading limit of one million tokens, which means it can handle very long conversations and large documents in a single session 1.
The model aims to help with hard problems in software engineering, finance, legal work, cybersecurity, and science. Google says it ranks number one on the Vals Index, an independent ranking that measures how well AI models perform in critical economic sectors 1.
From the official announcement
Speeding up code and science
Inside Google, thousands of Googlers are highlighting the model's strengths in specialized coding tasks, conducting deeper research, and writing quality. Some teams are using it through a tool called Antigravity. One team working on the Chrome browser used the model to untangle a ~6 million line folder of old code inside Chromium's 40 million-line codebase 1. By piping multiple clean code changes at a time, Argon helped finish an architectural refactor across hundreds of files in weeks instead of months .
The results will let webpages move smoothly between tabs, side panels, and floating windows without reloading, which should eliminate a whole category of browser crashes 1.
Another team behind Google's open-source Agent Development Kit gave Argon a rulebook and a one-paragraph goal. Working overnight on its own, the model found ways to cut preparation time by forty percent and speed up cold starts by twenty-one percent 1.
Breaking ground in science
Argon is also helping researchers in quantum computing, nuclear fusion, and life sciences. Scientists used it to optimize xTB, an open-source package for quantum chemistry, making it up to two-point-three times faster at predicting molecular structures 1. The changes were submitted to the official repository on GitHub .
Google also used Argon to make its open Gemma models nearly three times as fast on a specific type of processor, cutting a process that would normally take months down to just a few days 1.
How to get started
Argon is available through Google's API in AI Studio and the Gemini platform, as well as through Google Antigravity for Ultra subscribers and the Gemini app with AI Mode in Search 1. The price is two dollars per million input tokens and ten dollars per million output tokens at launch, with a ninety-five percent discount on cached input tokens .
Corrections and updates
- Corrected on 21 October 2026. An earlier version said the Chrome team untangled a sixty-million-line folder; the source states it was a roughly six-million-line legacy folder within a 40-million-line codebase. An earlier version said thousands of workers were using Argon specifically through Antigravity; the model is also accessible via the API, AI Studio, and the Gemini app, and the updated version correctly reflects that some teams use Antigravity while many others use it for coding, research, and writing tasks.
Key facts
- Gemini 4 Argon has a one-million-token limit for deep, multi-step problem solving
- Chrome team finished months of refactoring work in weeks using Argon
- Argon made an open-source quantum chemistry tool up to 2.3x faster
- Introductory pricing is $2 per million input tokens and $10 per million output tokens
Sources
Watch
If this story concerns you or your company, or uses something that belongs to you, ask for a correction or for it to be taken down. How requests work
Keep reading
Apps and productsGoogle opens TorchTPU, a native PyTorch backend for cloud TPUsHerald of AIhttps://github.com/soumith, CC BY-SA 4.0, via Wikimedia Commons
Google opens TorchTPU, a native PyTorch backend for cloud TPUs
A new open-source stack lets developers run PyTorch models directly on Google's TPU chips, replacing the older PyTorch/XLA system.
Business and rulesMirendil raises $1 billion to use AI agents for faster model developmentHerald of AI
Mirendil raises $1 billion to use AI agents for faster model development
A new startup founded by former Google DeepMind and Anthropic researchers says its automated platform lets small teams build large AI models quickly and cheaply.
New modelsGoogle’s Unreleased Gemini 4 Carbon Matches Anthropic’s Opus 5.5 on CodingHerald of AI
Google’s Unreleased Gemini 4 Carbon Matches Anthropic’s Opus 5.5 on Coding
An internal testing variant is matching the best coding model from a rival, and it hasn’t even launched yet.