Verified90% confidenceFactExact time

The Gemini series jointly models multiple modalities from the training stage, rather than appending a visual encoder as solutions like GPT-4V do.

7

Sources

90%

Confidence

Long-term

Relevance

5/29/2026

First Seen

Sources

Gemini Omni: How Powerful Is AI Video Generation That Understands Physics?

twitterGeminiApp5/29/2026

Related Entities

Geminiproduct GPT-4Vmodel Googlecompany OpenAIcompany

Related Claims

VerifiedGemini adopts a native multimodal architecture, jointly training the Transformer with mixed multimodal data from the pre-training stage, with tokens of different modalities processed together in the same sequence.78% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/179

API

curl https://kongchang.com/api/v1/knowledge/claims/179

MCP

get_claim(id=179)