Current AI’s first collaborative build
Most AI systems today are built behind closed doors and trained on Western languages. The new frontier is embodied AI: glasses, robots, pencils, speakers… Products that are entering our homes and bodies need to be open, personal, and multilingual. Current AI's first prototype — built with India's government-backed Bhashini and unveiled at the India AI Impact Summit on February 20th, 2026 — is a direct answer to that challenge.
What the device does
The handheld device can describe its surroundings via text or audio across the 22 official languages of India. It is entirely self-contained, runs AI models locally, and requires no internet connection.
It features a camera, screen, microphone, and speaker, running three AI models simultaneously: a vision model, Bhashini's automatic speech recognition and machine translation model, and a text-to-speech model. Together, they offer a powerful foundation for builders looking to create their own AI devices.
To enable others to examine, modify, and build on it, the device uses open-source software and hardware, and its full development instructions and code have already been released publicly. We want to give a shout-out to the open-source community that has been a huge source of inspiration for this build, particularly Marco Gerber’s Hear the world project.
Specifications
- 64 TFLOP of AI compute power
- 8GB of RAM
- 128GB of storage
- Battery: 15Wh
- Estimated 1-3 Hours of runtime
- Camera: 8MP, wide-angle
- Vibration motor for Haptic Feedback
- Dual-Band Wifi for software updates and development
- Otherwise fully-operational offline
- 2.8” Color TFT display
Model support out of the box
Instacart
Vision
- Ministral-3
- Qwen3-vl
- moondream
Speech recognition
- Vosk
- Bhashini
Instacart
Neural Model Translation
- Bhashini
Norrsken Foundation
Text to Speech
- Piper
- Bhashini
Pleias
A different vision for AI
“Our vision is to build a collaborative, collective, and global vision for AI, not one controlled by any single country or company”, says Current AI CEO Ayah Bdeir. “We will do so by partnering with mission-aligned organisations and creators to stitch together the full stack of public interest AI, from hardware to applications, and catalysing a movement for resilient public interest AI.”
What’s next
Current AI and Bhashini are launching a Global Innovation Challenge, inviting communities to imagine what this device could become and explore how open, embodied AI can enable many different futures—futures shaped by local needs, cultural contexts, and individual imagination—with complete privacy and control.
