已过期50% 置信事实精确时间
In the DeepSeek V4 + MIMO 2.5 composite architecture, when user input includes an image, DeepSeek V4 automatically triggers a vision tool call, letting MIMO 2.5 first describe the image before passing the description back to DeepSeek for reasoning.
1
来源数
50%
置信度
中期 (~90 天)
时效性
2026/7/2
首次发现
此条事实已过期，仅作历史留存，不再代表当前状态
有效期至：2026/9/30(已过期)
来源
1FlowBase in Practice: Adding Vision Tools to DeepSeek V4 for Multimodal Capabilities
bilibili老文-taichuy2026/6/21
相关事实
待验证DeepSeek-OCR是一个理解图片并识别其中文字的视觉模型，需使用Unsloth的FastVisionModel类加载69% 相似部分验证DeepSeek当前主力模型暂不支持图片识别（多模态），尚未开放图像理解的API接口67% 相似待验证DeepSeek V4 supports Function Calling, which is the technical prerequisite that enables it to proactively invoke external vision tools.66% 相似待验证DeepSeek团队在DeepSeek v4 flash上进行微调，并研发了自己的视觉编码器DeepSeek ViT66% 相似待验证Setting Reasoning Effort to Ultra in DeepSeek ZOI causes the DeepSeek model to perform noticeably better, generating frontend pages with complete dynamic effects and layout design64% 相似
引用此条事实
Stable URI
https://kongchang.com/claim/42729API
curl https://kongchang.com/api/v1/knowledge/claims/42729MCP
get_claim(id=42729)