English text is being prepared.

Google Releases EmbeddingGemma 2 Paper: A 744M-Parameter On-Device Multimodal Embedding Model

Google has released the EmbeddingGemma 2 paper, an Apache 2.0-licensed on-device multimodal embedding model that maps text, code, images, audio, and video into the same vector space. The model is built on Gemma 4 and tuned through contrastive training by learning from the larger Gemini Embedding 2; with 744M parameters, it surpasses models 2 to 12 times larger on multiple tasks. On the Pixel 11 Pro, the quantized text weights occupy about 191MB of active memory, while the full multimodal model takes about 567MB; the context window is expanded 4 times to 8K tokens, accommodating about 5.5 minutes of audio, 29 images, or 58 video frames; directly embedding speech scores 69.54 on the MSEB spoken query retrieval benchmark, higher than the 57.59 of transcribing first and then retrieving, and code retrieval on MTEB Code improves from 68.76 in EmbeddingGemma 1 to 78.68.

Iluvatar CoreX First Live Demonstration of TianGai 300 and On-Device Terminal TY1200

Iluvatar CoreX held the first live demonstration of its general-purpose GPU TianGai 300 and the on-device compute terminal TY1200, and previewed the next-generation on-device product TY2000 with a nominal compute of 600T. The TianGai 300 has 144GB of memory, delivering Attention performance 1.05 times that of the H200 under a 64K long context and MoE throughput 1.15 times that of the H200 on the GLM-5.3 model; the TY1200 offers 300 TOPS INT8 dense compute, with measured TPS surpassing NVIDIA's Thor T5000. The TianGai 300 IXL144 supernode integrates 144 GPUs in a single rack and aggregates about 10.1TB of memory, reducing the cost per million effective tokens by about 40% compared with mainstream international solutions.

Sentient Launches Open-Source On-Device AI Model Sentient OS

Sentient has released the open-source AI model Sentient OS, which can collect context from local files and messages and execute tasks on the user's computer. Its Sidekick feature allows users to hand off tasks such as open forms or emails on the screen for it to complete using its own cursor. What it learns is stored locally as plain-text files that users can open and edit; the first 1,000 users to enter FINAL_1000 in the app can use it free for life.

Sentient OS launches on-device AI: at night, a local LLM reads messages and files and writes them into markdown

Sentient OS (YC F26) has released on-device AI: at night, a local LLM reads the user's messages, files, and notes, and writes what it learns into a plain markdown folder on the computer, which the user can open, edit, or delete. During the day, Double Tap and Sidekick read these files to draft replies and complete tasks in apps, and only heavier tasks call the user-selected cloud model.

Interview with Bold Security Co-founder and CEO Nati Hazut: Using On-Device Small Models to Safeguard Enterprise Data

In the interview, Bold Security Co-founder and CEO Nati Hazut explained that the company uses small language models running directly on devices to identify sensitive data and how it is used, covering browsers, desktop applications, and AI agents. He pointed out that desktop AI applications and agents bypass the visibility of network detection, CASB, and DSPM, becoming new channels for data exfiltration; an on-device approach can intervene in real time before data leaves the enterprise's control.

CAICT, together with Huawei, vivo, StepFun and others, releases the "Device-Side Intelligent Industry Development Report (2026)"

The Institute of Artificial Intelligence of CAICT, together with the Device-Side Intelligence Group of the Artificial Intelligence Industry Development Alliance of China, and Huawei, ZTE, vivo, Haier, StepFun, ModelBest, Dahua Technology and other organizations, released the "Device-Side Intelligent Industry Development Report (2026)". The report defines device-side intelligence as a fused intelligence form that is primarily local and features device-cloud collaboration, focusing on key technologies such as model lightweighting, heterogeneous hardware integration, foundational software, agents, device-cloud collaboration, and security and trustworthiness. The report also outlines native product forms such as personal intelligence, spatial intelligence, mobile intelligence, industry intelligence nodes and embodied intelligence, as well as application scenarios such as consumer entertainment, smart home, smart office, smart mobility and smart health.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

Gradium open-sources the Phonon on-device TTS model suite, with all models under 50m parameters

Gradium open-sources the Phonon on-device speech synthesis model suite, with all models under 50m parameters, focused on multilingual naturalness and accuracy, and claims to offer the broadest inference support to date. Phonon-31m is released under the Apache 2.0 license. Matt Turck reposted this news.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.