Introducing the LENZO Model Zoo
Language models and object detection, verified on VPK180
We’ve added a Model Zoo to the LENZO documentation: a growing reference of models verified on the VPK180 TokenProcessor.
It gives developers a starting point for exploring workloads on TokenProcessor, bringing the model list, tested configurations and verification details together in one place.
Language and vision workloads
The collection currently spans English and Japanese text generation, alongside object detection. It includes language-model families such as Qwen, Llama, Gemma and Phi, Japanese models from LLM-jp and Sarashina, and YOLOv3-tiny for vision.
Models are run with matrix multiplications offloaded to TokenProcessor and their outputs checked against CPU-only runs under the documented test conditions. The Model Zoo explains how these functional checks are performed, along with model-specific configuration notes.
A reference that grows with the platform
We’ll continue adding models over time. The documentation is the place to find the latest coverage, quantization requirements and any compatibility caveats, with links to tutorials for running language models and comparing object detection on the CPU and TokenProcessor.
Whether you’re evaluating a familiar model family or preparing to try your own workload, start with the Model Zoo and follow the linked guides.
