Gemini 4 Argon: Built to solve your hardest problems

2026-10-14 · Google DeepMind

Gemini 4 Argon: Built to Solve Your Hardest Problems

On October 13, 2026, Google DeepMind and Chief AI Architect Koray Kavukcuoglu announced the global availability of Gemini 4 Argon, the latest frontier model designed for developers, enterprises, scientists, and consumers. The model is specifically built to handle deep work in complex, real-world tasks.

Availability and Pricing

Gemini 4 Argon is now accessible through multiple channels:

  • Developers and Enterprises: Via the API in AI Studio and the Gemini platform.
  • Consumers: Available for Google AI Ultra subscribers in Google Antigravity, and for both Google AI Ultra and Pro subscribers in the Gemini app and AI Mode in Search.

The model launches with an introductory pricing of $2 per million input tokens and $10 per million output tokens, with cached input tokens offered at a 95% discount.

Core Strengths and Benchmark Performance

Argon demonstrates exceptional depth and precision across multi-step tasks in domains such as coding, finance, legal, cybersecurity, and science. It outperforms other frontier models on the Vals Index, ranking #1 in measuring model performance across critical economic sectors including finance, coding, legal, and tax work.

Teams across Google, enterprises, and the research community have already leveraged Argon to achieve significant impact:

  • Solving major engineering problems like large codebase refactoring.
  • Optimizing business workflows in legal, finance, and science.
  • Finding and patching critical cybersecurity vulnerabilities.
  • Enabling breakthroughs in quantum computing, nuclear fusion, and life sciences.

Solving Complex Engineering Tasks

Google has utilized Argon on its most challenging engineering problems. Recently, software engineers have shipped high-quality code at record speeds. Using Argon in Antigravity increased agentic code submissions by over 35% and reduced code rollbacks by nearly a third.

Untangling Browser Feature Code in Chrome

Argon assisted the Chrome team in separating core webpage handling from a roughly 6 million line legacy folder within Chromium’s 40 million-line codebase. By piping multiple clean code changes simultaneously, Argon completed an architectural refactor across hundreds of files in weeks rather than months. This will allow live webpages to move fluidly between tabs, side panels, and floating windows without reloading, eliminating a category of browser crashes for billions of users.

Optimizing Python Codebases

Argon helped the team behind Google's open-source Agent Development Kit (ADK) push performance engineering further. Given a rulebook and a one-paragraph goal, Argon worked autonomously overnight to surface optimizations that cut in-memory session prep time by 40% and sped up cold imports by 21%.

Quantum Chemistry Package Optimization

Using Argon, researchers made xTB, a highly-cited open-source quantum chemistry package, up to 2.3x faster at predicting molecular structures and 1.8x faster at conformer search. These changes have been contributed to the official xTB GitHub repository.

Accelerating Gemma Models

Using Argon, Google achieved a 2.85x end-to-end inference speedup on an A100 via vLLM for Gemma 4 E2B, accelerating throughput from ~37,900 to over 108,000 tokens per second with 100% token parity and zero quality drop in just a few days.

Delivering Complex 3D Visualizations

Argon introduces major improvements in spatial reasoning, mathematical precision, and visual coding compared to previous Gemini models. Using slash commands and multi-agent loops in Antigravity, Googlers are prototyping complex 3D visualizations and interactive games in vanilla code that render directly in browsers.

Key examples include:

  • Pagoda Model: Argon autonomously sculpted a multi-building temple complex with photorealistic architecture and lighting in Blender.
  • Desk Racing Game: Argon transformed pictures of a real desk into a playable racing game, utilizing Blender MCP, Nano Banana textures, and Three.js for physics and audio.
  • Procedural 3D Game: Argon wrote Python scripts to procedurally sculpt and rig a 3D game, wiring it with GLSL shaders, shatter physics, and a single-player AI.
  • Ray-Tracing Engine: With zero external libraries, Argon wrote a pure Python 3D ray-tracing engine, producing a narrated video using Gemini 3.8 TTS and Lyria 3.5 music.
  • Gothic Architecture Scene: Starting from a single image, Argon autonomously drove Blender MCP to sculpt Gothic architecture and stage audio-reactive lighting.

Source