Google 官方介绍 Gemini 4 Argon:面向复杂编码与 Agent 任务
Google 官方账号称 Argon(Gemini 4)用于解决 Chromium 遗留代码与 Gemma 提速等复杂编码/Agent 问题,并开放研究者与网络安全者申请免费高配额。
From untangling legacy Chromium code to speeding up our @GoogleGemma models, Argon is built to solve highly complex coding and agentic problems at Google and beyond. We’re also opening applications for researchers and cyber defenders: 🔹 Scientists & researchers: Apply for a chance to get @Antigravity’s highest usage tier free for 6 months (subject to eligibility & selection) → http://goo.gle/4z0OCrZ 🔹 Cybersecurity defenders: Apply to join the Fairwind Program → http://goo.gle/3TUNMOj Learn more about Gemini 4 Argon ↓ http://goo.gle/4jfSznG
Gemini 4 Argon is also speeding up our @GoogleGemma open models — making them nearly three times as fast. In just a few days, Argon optimized Gemma 4 E2B to achieve a 2.85x inference speedup on an A100 via vLLM, jumping from ~37,900 to over 108,000 tokens per second with 100% token parity and zero quality drop. Here’s that speedup in action, generating 64 live Python simulations at once ↓ https://x.com/osanseviero/status/2110096066293207535/video/1
Software engineers at Google using Argon in @Antigravity are shipping high-quality code faster than ever. Over the past few weeks, agentic code submissions increased by more than 35% — while cutting code rollbacks by nearly a third. Read more about the impact of Gemini 4 Argon in Google Antigravity → http://goo.gle/4hPJbGo
In @GoogleChrome, Gemini 4 Argon helped our team separate core webpage handling from a 6-million-line legacy folder of browser features in Chromium’s 40-million-line codebase — completing months of work in just weeks. The result for Chrome's billions of users: eliminating an entire category of browser crashes, and letting live webpages move fluidly between tabs, side panels, and floating windows without reloading.
Teams across Google put Gemini 4 Argon to work on our hardest engineering problems. Here’s what happened 🧵
For Google's open-source Agent Development Kit (ADK), the team gave Argon a skills rulebook and one goal: eliminate framework overhead without breaking existing apps. Working autonomously overnight, Argon surfaced optimizations that cut in-memory session prep time by 40% in long sessions and sped up cold imports by 21%. See it live in google/adk-python → http://goo.gle/4eAtLDA