[LABS]
[MODELS]
Open weights
7B Vision Language Model
Reka Edge
Frontier-level visual intelligence, optimised for deployment. The fastest VLM in the 7B–8B class — built for real-world physical AI systems that can't wait for the cloud.
2.4×
faster avg. latency
7B
total parameters
64
tokens per image tile
HF + vLLM
easy deployment
Robotics & Physical AI
Sub-second spatial intelligence for autonomous robots — tool localisation, bounding boxes, and agentic motor control at the edge.
Sub-second latency
Bounding box grounding
Agentic tool use
Automotive & In-cabin AI
Privacy-first multimodal cabin agents running entirely offline on existing vehicle SoCs — no cloud roundtrip.
Offline-first
Multi-camera
ADAS integration
AR Wearables & Field AI
Hands-free visual Q&A for frontline workers in remote locations — 64 tokens per image tile preserves battery and device thermals.
Thermal-efficient
ffline visual Q&A
HUD integration
Real-time video & media
Content highlight generation, surveillance, and drone-based object detection — all at the edge without infrastructure overhead.
Video understanding
Object detection
Drone applications
[NEWS]
[PEOPLE]
A superstar roster of 50+ multimodal AI researchers and engineers — including pioneers behind models like Gemini and VEO.
+43
more researchers
and engineers
Full API reference, model cards, quickstart guides, and integration examples — everything you need to deploy Reka models in your stack.