Unverified50% confidenceFactExact time
DeepSeek-V-Flash-Vision-XP在获得多模态能力的同时完整保留了V-Flash原有的纯文本推理水准
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/10/2026
First Seen
Valid until: 12/9/2026
Sources
Related Entities
Related Claims
VerifiedDeepSeek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76% similarUnverifiedV4-Flash-Vision-Exp在纯文本Agent能力上与V4-Flash基本持平69% similarUnverified据B站UP主九天实测，DeepSeek V4 Flash搭配DSH V0.1.1的原生视觉能力可产出前端页面、HTML版PPT，并一句话生成接近商业级水准的宣传短片68% 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% similarUnverifiedDeepSeek V4 Pro不具备视觉理解能力，无法看懂截图，改为读取游戏内部状态和画布像素来判断渲染情况64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/890122API
curl https://kongchang.com/api/v1/knowledge/claims/890122MCP
get_claim(id=890122)