on-device AI
Tether Opens BotPsy-0.8B Weights for On-Device Agents
Tether AI Research has released BotPsy-0.8B, an on-device agent model built on Qwen3.5-0.8B, with Apache 2.0 weights. Benchmarks are Tether’s own claims.
Latest
Block Says Its Dessa Voice Model Learned to Talk Without Any Text
Block's Dessa is a full-duplex speech-to-speech model trained only on human conversation. It runs on a MacBook Pro; early access is for safety researchers.
iFlytek Says Its 293B Spark X2.6 Was Trained Only on Chinese Chips
iFlytek says its 293bn-parameter Spark X2.6 model was trained entirely on Chinese chips. It gave no benchmark scores. Only a 4B version has open weights.
BrainChip Launches Sequora, Chip Designs for Running Language Models on Battery-Powered Devices
BrainChip's Sequora is chip IP for running language models on battery-powered devices, using a state-space design it says keeps memory use fixed.
Alibaba Starts Selling Its Qwen Glasses N1 and N1 Pro AI Glasses in China
Alibaba has started selling its Qwen Glasses N1 and N1 Pro in China. Reported starting prices differ by outlet; Himax says its chip runs at 5 milliwatts.
Hugging Face Launches Llama Windows, a Free App That Runs Open AI Models From Your Windows 11 Tray
Hugging Face's Llama Windows runs open AI models on your Windows 11 PC from the tray, built on llama.cpp. Free, open source, no terminal or admin rights.
Google DeepMind and Singapore Build a Scam Detector That Runs on Your Phone
Google DeepMind and GovTech Singapore have built a Gemma-based scam detector that runs on phones and warns users. No launch date has been given yet.
Liquid AI Releases LFM2.5-24B-A2B, an Open-Weight Model for Running AI Agents on Your Own Device
Liquid AI's open-weight LFM2.5-24B-A2B has 24B parameters but uses only 2B at a time, built to run AI agents fast and privately on your own device.
Google Open-Sources ML Drift, the GPU Engine Inside LiteRT, Under Apache 2.0
Google has open-sourced ML Drift, the GPU engine inside LiteRT, under Apache 2.0. The 40%, 30% and 12% figures are Google's own claims; the old TFLite GPU delegate gets no new features.
NVIDIA Jetson Demo Runs a Gemma 4 Vision Kiosk On-Device
An NVIDIA tutorial shows OpenAI's Codex coding agent turning a Jetson Orin Nano into a self-starting kiosk that captions a live webcam feed with Google's small Gemma 4 E2B model, running on the device. What the demo shows, what "No Mac required" means, and why local inference matters.
Timeline
Every story in on-device AI, grouped by month.
November 2026
- 12 Nov 2026: Tether Opens BotPsy-0.8B Weights for On-Device AgentsTether AI Research has released BotPsy-0.8B, an on-device agent model built on Qwen3.5-0.8B, with Apache 2.0 weights. Benchmarks are Tether’s own claims.
October 2026
- 27 Oct 2026: Liquid AI Releases LFM3, Six Small Open-Weight Models for Any DeviceLiquid AI has released LFM3, six open-weight 2B-class models for text, vision, audio and decisions that it says run on phones and fine-tune on…
- 27 Oct 2026: Opera Puts Local AI in Its Browser, Giving Sites 3 Opt-In WebAI APIsOpera has begun shipping local AI models in Opera One as an opt-in WebAI preview, so websites and extensions can run AI on the…
- 27 Oct 2026: Apple Unveils M6 MacBook Pro and iMac a Day Before Its LA EventApple has announced a 14-inch MacBook Pro and an iMac with its M6 chip, from £1,999 and £1,499 in the UK, with pre-orders from…
- 24 Oct 2026: NVIDIA's Nemotron 3.5 Bolt E4B Packs an Agent Model Into About 4GBNVIDIA's open Nemotron 3.5 Bolt E4B, built on Gemma 4 E4B, runs voice and device agents in about 4GB. Its benchmarks are NVIDIA's, and…
- 23 Oct 2026: Block Says Its Dessa Voice Model Learned to Talk Without Any TextBlock's Dessa is a full-duplex speech-to-speech model trained only on human conversation. It runs on a MacBook Pro; early access is for safety researchers.
- 22 Oct 2026: iFlytek Says Its 293B Spark X2.6 Was Trained Only on Chinese ChipsiFlytek says its 293bn-parameter Spark X2.6 model was trained entirely on Chinese chips. It gave no benchmark scores. Only a 4B version has open…
- 14 Oct 2026: BrainChip Launches Sequora, Chip Designs for Running Language Models on Battery-Powered DevicesBrainChip's Sequora is chip IP for running language models on battery-powered devices, using a state-space design it says keeps memory use fixed.
- 14 Oct 2026: Alibaba Starts Selling Its Qwen Glasses N1 and N1 Pro AI Glasses in ChinaAlibaba has started selling its Qwen Glasses N1 and N1 Pro in China. Reported starting prices differ by outlet; Himax says its chip runs…
- 14 Oct 2026: Hugging Face Launches Llama Windows, a Free App That Runs Open AI Models From Your Windows 11 TrayHugging Face's Llama Windows runs open AI models on your Windows 11 PC from the tray, built on llama.cpp. Free, open source, no terminal…
- 14 Oct 2026: Google DeepMind and Singapore Build a Scam Detector That Runs on Your PhoneGoogle DeepMind and GovTech Singapore have built a Gemma-based scam detector that runs on phones and warns users. No launch date has been given…
- 13 Oct 2026: Liquid AI Releases LFM2.5-24B-A2B, an Open-Weight Model for Running AI Agents on Your Own DeviceLiquid AI's open-weight LFM2.5-24B-A2B has 24B parameters but uses only 2B at a time, built to run AI agents fast and privately on your…
Load more Showing 12 of 14
Related stories
Latest from topics connected to on-device AI. See all with connected topics
Latest on TSN
Explore the TSN Atlas: Every topic we cover, mapped. Click a bubble to dive in.
