Frequently Asked Questions about ZETTA AI Box

No. ZETTA AI Box is designed to operate entirely on-premise and can function fully offline, without an Internet connection or any cloud services. It can also be connected to a secure private network when required. This architecture makes it particularly suitable for sensitive, classified, mission-critical, or disconnected environments.

The ZETTA AI Agent is the agentic layer that works on top of every AI engine in the box. Analysts ask questions in natural language, and the agent searches all AI-processed evidence, identifies relationships between documents, audio, images and video, and generates structured, investigation-ready reports, entirely on the box with no cloud dependencies.

ZETTA AI Box provides multimodal AI processing for documents, scanned materials, audio, images, and video. Capabilities include OCR and handwriting recognition, speech-to-text and speaker diarization, machine translation, entity extraction, classification, summarization, object detection, key-frame extraction, video description, and metadata analysis. These capabilities can be combined within a single investigation or analytical workflow.

The core multimodal processing capabilities of ZETTA AI Box are powered by ZETTA’s proprietary AI Engines, optimized for tasks such as speech, language, document, image, and video analysis. These are complemented by selected open-weight large language models used for advanced analytical tasks, including summarization, natural-language interaction, evidence analysis, and report generation.

Yes. ZETTA AI Box is designed for multilingual environments and supports speech processing, machine translation, and text analytics across up to 100 languages. It can automatically identify languages, transcribe multilingual audio, translate documents and conversations, and analyze information across languages within the same investigation workflow.

Yes. ZETTA AI Box incorporates human-in-the-loop workflows that allow analysts to review, validate, and post-edit AI-generated outputs. Access can be managed at user and project level, ensuring appropriate control over sensitive content. Reviewed and approved datasets can also be used to adapt or fine-tune selected AI capabilities directly on the Box through ZETTA Factory, ZETTA’s no-code AI training and adaptation environment.

ZETTA AI Box supports three main deployment modes. It can operate as a standalone workstation connected directly to an HDMI display, keyboard, and mouse; as a network appliance accessible by multiple authorized users over an internal private network; or in Access Point mode, allowing users to connect directly from supported mobile devices such as smartphones and tablets. This flexibility enables deployment in offices, investigation rooms, mobile units, and isolated environments.

ZETTA AI Box is built on systems powered by the NVIDIA® GB10 Grace Blackwell Superchip, delivering up to 1 PFLOP of FP4 AI performance and 128 GB of coherent unified memory. This enables demanding AI workloads to run locally on a compact system without dependence on external cloud infrastructure. Supported hardware platforms may include systems such as the Dell Pro Max with GB10, among other compatible GB10-based devices.

Software and model updates can be delivered on verified offline media, following your organization’s security procedures.

Yes. Test ZETTA AI Box on your own type of data, on-site.

Directly from ZETTA or through our partners in your country. Public sector buyers: we support tenders and NCAGE-registered procurement.