Uploaded September 2026 | Updated November 2026, 2 days ago
Mateusz, who leads our internal AI team, asked Reachy Mini where on-device AI actually stands right now. The robot hears with Whisper and speaks with Kokoro, both running locally, while Claude handles the conversation. 🤖
Mobile is our home turf, so of course we have to bring it up: React Native ExecuTorch is where a lot of our on-device work ends up, and it is the fastest way to run the same models in your own app.
➡️ executorch.swmansion.com
Follow us ⬇️
Software Mansion: https://x.com/swmansion
Mateusz: https://x.com/mkopcins
Timestamps
0:00 Meeting Reachy
0:18 Where on-device AI stands in 2026
0:47 What small models handle well
1:05 Cutting cloud costs by routing work to devices
1:32 Vision and small multimodal models
1:53 Running YOLO on device or in the cloud
2:15 Libraries: React Native ExecuTorch, LiteRT, ONNX Runtime Mobile, TypeGPU
2:48 Three apps to try on your phone today
3:09 Wrap-up
#huggingface #ondeviceai #privacy #mobileai #semanticsearch #llms #aidevelopment #softwaredevelopment #pytorch #edgeai #pollenrobotics #localai #ondeviceai #offlineai
Mateusz, who leads our internal AI team, asked Reachy Mini where on-device AI actually stands right now. The robot hears with Whisper and speaks with Kokoro, both running locally, while Claude handles the conversation. 🤖
Mobile is our home turf, so of course we have to bring it up: React Native ExecuTorch is where a lot of our on-device work ends up, and it is the fastest way to run the same models in your own app.
➡️ executorch.swmansion.com
Follow us ⬇️
Software Mansion: https://x.com/swmansion
Mateusz: https://x.com/mkopcins
Timestamps
0:00 Meeting Reachy
0:18 Where on-device AI stands in 2026
0:47 What small models handle well
1:05 Cutting cloud costs by routing work to devices
1:32 Vision and small multimodal models
1:53 Running YOLO on device or in the cloud
2:15 Libraries: React Native ExecuTorch, LiteRT, ONNX Runtime Mobile, TypeGPU
2:48 Three apps to try on your phone today
3:09 Wrap-up
#huggingface #ondeviceai #privacy #mobileai #semanticsearch #llms #aidevelopment #softwaredevelopment #pytorch #edgeai #pollenrobotics #localai #ondeviceai #offlineai










