A mid-sized multimodal model from the Qwen 3 family, capable of processing both text and image inputs to produce text outputs. It carries the NVFP4 quantization with a BF16 language model head, a configuration that trades some numerical precision for reduced memory footprint while preserving output quality at the final layer. Published by RadixArk under an open Apache 2.0 license, it reflects a practical engineering approach to deploying capable models on constrained hardware.