Skip to content
IFAI

On-device AI

Industry tag622 stories

  1. Open sourceReach 69

    Mistral drops Mistral 3 family under Apache 2.0

    Mistral AI just unleashed Mistral 3, an entire family of multimodal models released under a fully permissive Apache 2.0 license. The drop includes Mistral Large 3, a massive 675B-parameter MoE model (41B active), alongside edge-friendly Ministral 3 variants spanning 3B, 8B, and 14B sizes with native vision support. Why it matters: developers get frontier-grade open weights they can actually modify and deploy commercially without restrictive licensing handcuffs.

  2. Open sourceReach 57

    NVIDIA gives self-driving cars a chain of thought

    NVIDIA just expanded its open-source footprint at NeurIPS with Alpamayo-R1, the first industry-scale open reasoning VLA model for autonomous driving. The model pairs chain-of-thought reasoning with path planning, helping vehicles handle messy edge cases like crowded intersections by breaking down situations step by step. Why it matters: bringing common-sense reasoning into physical AI could accelerate the path toward safe level 4 autonomy.

  4. IndustryReach 36

    Qualcomm brings top-tier AI agents to cheaper phones

    Qualcomm just introduced the Snapdragon 8 Gen 5, a lower-cost processor bringing advanced on-device AI features to sub-flagship Android phones. Powered by the company’s Oryon CPU, it supports agentic features like personal contextual suggestions while boosting CPU performance by 36% over the 8 Gen 3. The first devices, including the OnePlus 15R, arrive next month.

  5. Open sourceReach 49

    Microsoft's tiny AI agent can click and type for you

    Microsoft just open-sourced Fara-7B, a compact 7B-parameter model specifically built to take over keyboard and mouse actions. By reading raw browser screenshots instead of code trees, it automates routine web tasks like booking trips or comparing prices. Why it matters: It is small enough to run locally on Copilot+ PCs, delivering fast agentic workflows without uploading private screen data.

  6. Open sourceReach 46

    Hugging Face drops unified AI API for Apple platforms

    Hugging Face just launched AnyLanguageModel, an open-source Swift package that serves as a drop-in replacement for Apple's Foundation Models framework. By swapping a single import statement, developers can alternate between local engines like MLX and Core ML and cloud providers without rewriting glue code. It dramatically lowers the bar for running open-source models on Apple silicon.

  7. IndustryReach 33

    Rivian sets Autonomy and AI day for hands-free driving

    Rivian just announced an Autonomy and AI day for December 11, where it plans to preview its "Universal Hands Free" driving system. CEO RJ Scaringe recently completed a two-hour test drive across Palo Alto with a second-gen R1 driving itself. Rivian targets a controlled hands-off rollout by 2026, though specific pricing and launch timelines may remain under wraps.

  8. RoboticsReach 47

    MindOne puts an AI housekeeper brain into Unitree's G1

    Shenzhen startup MindOne Robotics just shared a viral demo showing a Unitree G1 humanoid tackling household chores like watering plants, wiping counters, and fetching packages. The company claims the entire video runs autonomously with zero teleoperation or speed-ups, sparking intense industry curiosity given humanoid robotics' notorious generalization hurdle.

  9. IndustryReach 63

    Even Realities drops the camera-free G2 smart glasses and R1 ring

    Even Realities just introduced the G2 smart glasses and R1 smart ring, prioritizing discreet assistance over sensory overload. The 36-gram G2 ditches cameras and speakers for clean optics and AI conversation tracking, while the R1 ring acts as a gesture controller and health tracker. Prices start at $599 and $249.

  10. AI toolsReach 35

    Google Messages turns chat photos into action figures

    Google just rolled out Remix in Google Messages, an in-app photo editing tool that uses Nano Banana, the Gemini image model. Users can long-press any image in a conversation to change hairstyles, redesign rooms, or transform pets into action figures without leaving the app. The feature lets friends remix photos back and forth, and is currently rolling out exclusively on Android for users 18 and older.

  11. Open sourceReach 35

    Baidu drops open-source ERNIE multimodal model that "thinks with images"

    Baidu just open-sourced ERNIE-4.5-VL-28B-A3B-Thinking under an Apache 2.0 license. The model boasts 28B total parameters but activates only 3B at runtime, featuring a dynamic zoom-in mechanism to inspect visual details. Baidu claims it beats Gemini 2.5 Pro and GPT-5-High on key visual reasoning benchmarks, though independent testing is still pending.

  12. Open sourceReach 46

    Weibo AI open-sources a $7,800 reasoning model

    Weibo AI just open-sourced VibeThinker-1.5B, a 1.5B-parameter reasoning model trained for just $7,800. On benchmarks like AIME24 and AIME25, it outscored the original DeepSeek R1 despite having a fraction of the parameters. It is tailored specifically for competitive math and coding problems in English.

  13. Open sourceReach 35

    Manifest AI drops attention from Qwen3 for just $4,000

    Manifest AI just unveiled Brumby-14B-Base, a retrained variant of Qwen3 that completely ditches the transformer’s signature attention mechanism. Instead, it uses a recurrent architecture called Power Retention, keeping per-token compute costs flat across long contexts. The team trained it on 32 H100s for just $4,000 to match baseline performance, though they admit the cost-saving trick only works by building on existing transformer weights.

  14. Open sourceReach 40

    PewDiePie built a 10-GPU rig to self-host massive AI models

    PewDiePie just went all-in on local AI, building a custom mini-datacenter with 10 GPUs—including modded 48GB RTX 4090s—to run massive open-source models offline. He created a custom interface called ChatOS, orchestrated a multi-bot "council," and plans to release his own self-hostable model soon. Why it matters: It shows local, privacy-first AI compute is rapidly hitting the mainstream enthusiast space.

  15. IndustryReach 34

    Musk's wild plan for idle Teslas

    Elon Musk just floated the idea of turning idle Teslas into a giant distributed inference fleet. He pitched that an eventual fleet of 100 million vehicles could pump out 100 gigawatts of AI compute when parked. Why it matters: It sounds like a massive decentralized data center on wheels, though convincing owners to burn extra battery and wear down chips remains a huge hurdle.

  16. AI toolsReach 41

    Microsoft Paint gets an AI Restyle button

    Microsoft just rolled out a new AI feature called Restyle to Windows 11 Insiders in Paint. It lets you instantly transform the artistic style of any canvas image by picking from preset styles in the Copilot menu. The catch: it requires a Microsoft account and only runs on Snapdragon-powered Copilot+ PCs.

  17. IndustryReach 50

    GM unveils eyes-off driving for Cadillac, adds Gemini AI

    GM just announced eyes-off highway driving is coming in 2028, debuting on the Cadillac Escalade IQ before expanding across its lineup. The setup combines Super Cruise with Cruise technology and LiDAR-based sensor fusion on a platform packing 35 times more AI performance. Drivers won't have to wait that long for smarter cabins, though: GM is rolling out Google Gemini-powered conversational AI to vehicles starting next year.

  18. AI toolsReach 32

    YouTube adds AI-generated Q&A stickers and Shorts limits

    YouTube just rolled out AI-powered smart Q&A stickers for live streams, letting creators automatically generate prompts to poll their audience when out of ideas. The feature is currently available in English on the mobile app. YouTube also introduced daily viewing time limits to curb endless Shorts scrolling.

  19. ResearchReach 36

    LabOS turns AI into an in-lab co-scientist with XR glasses

    LabOS just introduced an AI co-scientist that bridges digital reasoning with physical lab benches via XR glasses. The system tracks your hands and tools to catch wrong reagents or missed steps in real time. Both software and hardware are slated to be open-sourced, with access requests now open.

  20. IndustryReach 40

    Apple gives Vision Pro an M5 boost and a comfier band

    Apple just refreshed the Vision Pro with an M5 chip and a cushioned Dual Knit Band, starting at $3,499. The new silicon delivers up to 50% faster system AI tasks, renders 10% more display pixels, and bumps battery life to 2.5 hours. Why it matters: Apple is tackling the biggest pain points—comfort and horsepower—while keeping its premium spatial computing bet moving.