← All models

CogAgent-Chat

Zhipu AI (Z.ai) · 2023-12-15

CogAgent-Chat is Zhipu AI and Tsinghua University's original CogAgent checkpoint for GUI navigation, visual grounding, and multi-turn visual dialogue. The 18B-class multimodal model combines a language backbone with visual experts and a high-resolution cross-attention module for 1120×1120 images. It is distinct from CogAgent-VQA, and its weights use the CogVLM License rather than the repository code's Apache-2.0 license.

Benchmark scores