Nace AI just released a 51-page technical report on Drex, its open-weight model family designed to pick the best answer from fixed options. Trained on 1.15 million verified decisions, it delivers calibrated probabilities in just 183ms. Open-weight versions are available on Hugging Face alongside a web test console.
Cactus just dropped Mirage, a tiny 51.3M-parameter vision model compressed to roughly 22MB via 2-bit quantization. It handles scene description, object detection, OCR, segmentation masks, and spatial reasoning in a single pass—clocking 74ms on a Galaxy S26 NPU. It proves full-featured visual perception can comfortably live on device edge chips without touching the cloud.
Video Rebirth just released HyperFlow H3 Flash, an open ~14B distillation of MiniMax-H3 that generates synchronized 768p video and audio in 4 to 8 steps. The Q4 GGUF weight drops down to 9.1 GB to run on 24GB RTX GPUs, while 4x Blackwell B200s achieve playback-speed real-time generation. Why it matters: Cutting-edge, synchronized audiovisual models are finally breaking out of hyperscaler clusters and onto local creator hardware.
Open-source project hunch turns Google's EmbeddingGemma 2 into a local, reflex-speed decision engine running directly on commodity CPUs. It acts as a fast student that learns from frontier LLMs and takes over repetitive chores like tool trace verification in just 22ms. Basically, it lets you slash agent inference bills without needing a GPU.
Nous Research just opened a waitlist for its new Hermes Mobile app after initial beta spots filled up almost instantly at NousCon 2026. The team says it is actively working to free up more beta slots while polishing the agent app for a wider public rollout.
Apple just introduced always-on "Audio Intelligence" features in the watchOS 27.3 developer beta, bringing conversational AI directly to your wrist. A double-press of the Digital Crown pulls up a 15-second text rewind of what was just said, while Siri Recap automatically jots down key points and titles. Apple claims raw audio is processed strictly on dedicated hardware and never saved as a file.
Netflix just launched an experimental 'AI Discover' tab on its TV app, serving up themed picks based on your mood—like 'Something Light' or 'Quick Watches.' Powered by a new intelligence layer, it remembers titles you reject and tunes suggestions in real time. It marks a shift away from endless grid scrolling, rolling out first to English-language TV users in the U.S.
Researchers just open-sourced NitroFusion 2, slashing image diffusion down from 8 steps to just 1 or 2 steps. By introducing adversarial distillation directly into the multimodal text stream, it preserves fine details and legible typography even in a single pass. The weights are available on Hugging Face and run locally via ComfyUI.
Researchers from Tsinghua and UC Berkeley just open-sourced uni-robot-agent, a unified harness that lets open-weight models control robots in simulation and the real world. Running locally on Qwen3.8-Flash-Next, it scored 31.39 on RoboDojo, edging past GPT-6 Astra's baseline at roughly 1/20th the cost.
StudyOnTV just turned screen time management into a learning gate for Android TVs and tablets, forcing kids to answer questions before earning viewing minutes. It runs completely locally with no cloud accounts, and can plug into AI models like DeepSeek to auto-grade subjective answers.
Subvocal just raised $1.3M in pre-seed funding to build smart glasses that turn silent mouth movements into text. Using low-power radar-like radio frequency sensors and AI rather than cameras or microphones, the device tracks facial muscle shifts at over 200 words per minute. Pre-orders start at $499, targeting a 2027 rollout.
Google just fired up its Project Suncatcher prototype satellite, generating the first-ever AI response from a data center-grade TPU in orbit. Passing over Norway, the satellite ran Google's Gemma 4 model to beam back a space-themed haiku. Why it matters: Google is testing whether AI chips can survive extreme radiation and thermal swings, setting the stage for orbital data centers.
Edge0 just open-sourced Audio8-TTS-Realistic, a lightweight 601M-parameter text-to-speech model supporting 23 languages. It pulls off zero-shot voice cloning from just a few seconds of clean reference audio and weighs only 2.6GB under an Apache 2.0 license. Why it matters: you can now run fast, commercial-friendly voice cloning on consumer-grade GPUs without massive compute.
Pika just open-sourced Pika H3 Flex, a post-trained model built on MiniMax-H3 that generates videos with synchronized stereo audio in only six sampling steps. It handles 5-to-15-second clips across 11 aspect ratios, accepts up to 12 multimodal references, and runs on a single H100 GPU. Why it matters: professional creators get rapid, high-control iteration without waiting minutes for a render.
Sperid Labs just open-sourced Idol, a new model that turns a single 2D image into a standard 3D Gaussian splat with up to a million points. It handles background removal, accurately guesses the hidden backside, and caps out at just 7.2 GB of VRAM. Why it matters: High-fidelity 3D reconstruction is finally fast and lean enough for local desktop rigs.
Developer Nutlope just open-sourced Bygga, a web app that turns flat IKEA paper instructions into interactive, animated 3D guides. It covers 1,171 products completely offline, reads steps aloud, and listens for voice commands so you can say "next" without dropping your screwdriver. Unknown manuals can be scanned and drafted via Together AI.
AI 3D platform Hi3D just open-sourced its V1.0 model weights and code under a commercial-friendly license. Sized at just 7.74GB, it runs on consumer GPUs to turn single images into detailed meshes up to 1536³ resolution. It hands creators a solid local baseline, even though the flagship V3 remains closed-source.
Baidu just open-sourced Unlimited OCR 2, a 3B document-parsing model that can ingest thousands of pages in a single request. By revamping attention with SR-SWA, it keeps memory and latency stable over massive documents, automatically stitching together cross-page tables and split text. It easily fits on consumer GPUs, meaning you no longer have to chop books into pieces just to digitize them.
Block just unveiled Dessa, a real-time voice model trained entirely on human speech without relying on text or traditional pre-training. It can laugh, sigh, and interrupt naturally while running locally on consumer hardware. Public access is currently locked down, with testing limited to AI safety researchers.
Researchers from Tsinghua University just open-sourced Falcon, a lightweight model that isolates target voices in noisy rooms by tracking mouth movements. With just 0.5M trainable parameters, it drastically cuts latency compared to prior baselines. The accompanying Falcon Studio app runs on a laptop and feeds the clean audio straight into Zoom or Tencent Meeting.
