谷歌新前沿模型在编程与安全上超越 Opus 和 GPT-6 Astra

Ejaaz @cryptopunk7213原推发布于 2026.10.29

在 X 上打开原推

holyyy shit google is so damn back, new frontier model that beats fable, opus and GPT 6 astra at coding and cybersecurity 🫪 two observations: 1. they dropped this model so SOON after Argon. the rumors of google achieving RSI internally could be true. 2. remarkable how they’ve been able to catch up despite being so far behind earlier this year. money + data really helps. disclaimer: these might just be benchmaxxed scores so full review once i actually get access to it!

这条是引用着另一条写的：「Google DeepMind 官宣 Gemini 4 Carbon 为下一代前沿模型」