Unverified75% confidenceFactTime unknown

Gemini 2.5 Pro采用原生多模态架构，从训练阶段就融合文本、图像、音频和视频

1

Sources

75%

Confidence

Medium-term (~90 days)

Relevance

6/1/2026

First Seen

Valid until: 8/30/2026

Sources

SAP AI Core LLM Proxy：统一接入GPT-5等主流大模型的企业级代理方案

githubpjq

Related Entities

Gemini 2.5 Promodel Googlecompany

Related Claims

UnverifiedGemini从预训练阶段就将文本、图像、音频、视频数据混合输入，实现原生多模态训练88% similar UnverifiedGemini的图像生成能力基于原生多模态架构，模型在训练阶段同时学习了文本和图像的表征83% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/29428

API

curl https://kongchang.com/api/v1/knowledge/claims/29428

MCP

get_claim(id=29428)