Google releases Gemini 4 Argon with 1M output tokens
Google's new frontier model raises the output limit from 64K to 1M tokens, scores 77.9% on DeepSWE v1.1 and 68% on CWE-Bench, and is first available internally and to trusted cyber defenders.
Google is so back 🔥🔥 Google just announced Gemini 4 Argon, their new frontier model, and the numbers are absurd. - 1M output tokens: a massive jump from 64K. It can generate hundreds of thousands of tokens in a single run to reason through massive codebases and long workflows in one go. - Coding SOTA: sets a new high score on DeepSWE v1.1 at 77.9% for long-horizon software engineering. - Autonomous zero-day patching: ties for first place on CWE-Bench at 68%, finding and fixing critical software exploits autonomously. - Enterprise benchmarks: ranks #1 on the Vals Index across finance, legal, and tax, plus 51.3% on Zapier's AutomationBench. For now, Google is rolling it out internally and to trusted cyber defenders in their Fairwind Program alongside US government pre-release testing before opening access to paid API and Google AI Ultra users.