Unverified50% confidenceFactExact time
OpenBMB's MiniCPM-V 4.7 continues the combination of a SigLIP vision encoder plus a Qwen3.5 language model backbone, supports two visual downsampling modes, 4x and 16x, and adds canvas M-RoPE to lay images on a 2-D canvas to assign (temporal, height, width) positions to visual tokens.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/15/2026
First Seen
Valid until: 1/13/2027
Sources
Related Entities
Related Claims
UnverifiedQwen3 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70% similarUnverifiedMainstream large image generation models such as SD3, FLUX, and Qwen-Image-3.0 have broadly shifted toward an architecture that fuses LLM text encoders with DiT image backbones.70% similarUnverifiedLLaVA、InstructBLIP、Qwen-VL等主流多模态架构采用冻结或微调视觉编码器+线性投影+文本LLM的三段式结构68% similarUnverifiedQwen3.8-Flash-Next supports mixed text and image input, with a maximum context length of 262K tokens, and offers four adjustable reasoning intensity levels: None, Low, Medium, and Extra High.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1039065API
curl https://kongchang.com/api/v1/knowledge/claims/1039065MCP
get_claim(id=1039065)