GLM-4.6V (NovitaAI)

GLM-4.6V is a vision-language model from Zhipu AI, continuing the GLM series with a strong emphasis on visual reasoning alongside language understanding. According to Novita AI's documentation, the foundation version carries 106 billion parameters and is trained with a 128K-token context window, giving it room to handle lengthy documents, extended video-derived frame sequences, and other inputs that demand long-range coherence. Novita's model page frames GLM-4.6V as a "significant multimodal advancement" that achieves state-of-the-art visual understanding accuracy for models of its parameter scale, positioning it as a flagship option for developers who need both rich image comprehension and the conversational depth of a large language model.

What sets GLM-4.6V apart from earlier visual models is that function calling is wired directly into the architecture rather than bolted on through prompting or external orchestration. Both Novita's model page and its launch blog describe it as the first visual model to natively integrate Function Call capabilities, creating a direct bridge from what the model sees to tools it can invoke. That design choice makes GLM-4.6V well suited to multimodal agents that must read a screen, parse a chart, and then call APIs, as well as to document-processing pipelines and visual editing applications where perception has to translate into action in a single step.