Gemini Embedding 2：Google 的原生多模态向量模型，工程上怎么用？
Google 发布 Gemini Embedding 2（gemini-embedding-2-preview），把文本、图片、音频、视频、PDF 映射到同一向量空间。本文从工程实践角度拆解它的规格、维度选择与落地路径。
#Gemini
5 posts found.
Gemini 3.1 Pro：推理分数翻倍之后，我们该兴奋还是该冷静？
Google 发布 Gemini 3.1 Pro，ARC-AGI-2 推理分数从 31.1% 跃升至 77.1%，超越 Claude Opus 4.6 和 GPT-5.2。这篇文章拆解它到底强在哪、弱在哪、以及对开发者意味着什么。
Your AI Copilot in the Terminal: A Guide to the Gemini CLI
Forget context switching between your editor and a browser tab. The Gemini CLI brings Google's powerful AI directly into your terminal, supercharging your development workflow. This guide explores how to get started.
The "Biased Genius" with 91% Hallucination Rate: In-Depth Review of Gemini 3 Flash
Google Gemini 3 Flash performed amazingly in the Artificial Analysis evaluation: tied for first place with flagship models, 55% accuracy, but with a hallucination rate as high as 91%. This article deeply analyzes the strengths and pitfalls of this "biased genius".
Gemini 3震撼发布：AI智能体时代的真正黎明
深度解析Google Gemini 3.0的技术突破、行业影响以及AI智能体时代的来临