Gemini 4 Argon 多项 benchmark 反超 Opus 5.5 与 Astra
汇总 Argon 跑分：DeepSWE 77.9%、CWE-bench 68% 并列第一、AutomationBench 51.3% 第一，输出上限从 64K 提升到 1M token，将先向 Fairwind 计划与 Ultra 订阅者开放。
Google just crashed the Anthropic vs OpenAI fight with Gemini 4 Argon. it beats Opus 5.5 and GPT-6 Astra on DeepSWE, and it can write up to 1M tokens in a single answer. the numbers: • DeepSWE v1.1 → Argon 77.9%, Opus 5.5 74.2%, Astra 74.1% • CWE-bench v1 → 68%, tied for #1 • AutomationBench → 51.3%, #1 • output limit → 1M tokens, up from 64K who gets it first: → cyber defenders through Google's Fairwind program → then AI Ultra subscribers and paid API customers Anthropic led. OpenAI chased. Google just showed up with a better score. https://x.com/Google/status/2105388143902175529?s=20
这条是引用着另一条写的：x.com/google/status/2105388143902175529