Unverified50% confidenceReleaseExact time
DeepSeek将实验性多模态模型V4-Flash-Vision-EXP上线API平台，其文本能力与V4 Flash持平
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/25/2026
First Seen
Valid until: 11/23/2026
Sources
Related Entities
Related Claims
UnverifiedDeepSeek V4 Pro用Manim可视化二次函数仅耗时4分钟输出1080P视频，效果优于V4 Flash74% similarUnverifiedDeepSeek V4 Pro在3D跑酷游戏任务中加入了镜头跟随和音效功能，表现超出预期70% similarUnverifiedQwen3 27B是原生多模态模型，支持视觉输入，弥补了DeepSeek V-Flash无法处理图像的短板67% similarUnverifiedSTEP3.7 Flash具备端到端多模态实时交互能力，可同时处理文本、图像、音频、视频等多种输入模态66% similarUnverifiedFlash模型定位为轻量、高速、低成本，适合高并发请求、内容分类、摘要生成、客服问答、批量数据处理等场景63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/799682API
curl https://kongchang.com/api/v1/knowledge/claims/799682MCP
get_claim(id=799682)