MiniCPM-V 4.6

1.3B-parameter on-device vision-language model with 262K context window and 4x/16x mixed visual-token compression for iOS, Android, and HarmonyOS deployment. AA Intelligence Index v4.0: 13 — reportedly 19x more token-efficient than Qwen3.5 0.8B at comparable quality.

Continues the MiniCPM-V series' push on the edge-multimodal frontier alongside the omni-modal MiniCPM-o 2.6.