Rocket - Start building, Now!
Mini toolsReka Edge
By Reka AI
Reka Edge uses a ConvNeXt V2 vision encoder plus a transformer backbone trained from scratch, and is optimized for token-efficient visual processing to support high-definition images and streaming video without consuming excessive context. It targets physical-AI workloads with capabilities including image understanding, video analysis, object detection and localization (bounding boxes via a detection prompt format), and tool calling for agent workflows, with an emphasis on low latency and high throughput.
New
Multimodal
Gen 3
Overview
Reka Edge is Reka’s 7B vision-language model engineered for edge and physical-AI deployment, with strong visual reasoning, physical grounding, and agentic tool use.
Tools using Reka Edge
Last updated: March 12, 2026
