Reka Edge

Reka Edge is a robust 7B Vision Language Model optimized for real-world physical systems. It integrates a highly efficient ConvNeXt V2 encoder, enabling image processing with significantly fewer tokens. This design delivers sub-second latency for critical tasks such as real-time video analysis, precise object detection, and agentic tool use. Its architecture reduces latency by up to 2.4x on single requests, making it ideal for applications where speed and responsiveness are paramount. The model also offers a smaller memory footprint due to its efficient token usage, ensuring high performance even in resource-constrained environments. Designed for edge deployment, Reka Edge excels across various multimodal benchmarks, providing strong capabilities in video understanding, image comprehension, and sophisticated tool integration. It is suitable for diverse use cases including powering autonomous robots with spatial awareness, facilitating instant object manipulation, and generating media highlights. This model is perfect for engineering teams and developers building demanding physical systems, robotics, and real-time surveillance applications that require immediate and reliable visual intelligence.