Modal Blog7575NVIDIA has delivered its first Vera CPU server to Modal, an Arm-compatible CPU that, through emulation capabilities testing, has increased the throughput of the emulation capability sandbox by 2x, suitable for heterogeneous workloads with peak resource usage up to 10x the baseline, helping to reduce costs and accelerate reinforcement learning deployment.
Google Developers BlogLiteRT Tensor API is released, supporting developers to write and optimize models natively in C++ on edge devices, enabling cross-platform deployment (CPU/GPU/NPU) without script conversion. Developers can fully control the computational graph structure and directly load model weights, achieving zero-copy data transfer through shared accelerator buffers, while providing building blocks for AI coding agents.
Inception Labs BlogMercury 2 is the first reasoning diffusion language model, decoding over 1000 tokens per second on standard NVIDIA GPUs, reducing inference cost from 3 seconds of silence to 300 milliseconds.
AI as Normal TechnologyBeyond Automation: A new research direction in AI argues that AI should not be limited to automating human labor but should explore new directions for enhancing humans and promoting human-machine collaboration. This perspective advocates for AI research to focus on new issues such as transparent agents, process tracking, and human-machine interaction design to push AI beyond pure technological automation and achieve broader applications.
Google DeepMindGemini 4 Argon has achieved major breakthroughs in physical understanding, code generation, and visual semantic reasoning, achieving a pure reasoning score of 87% on the Embodied Reasoning (ERQA) benchmark test, surpassing all leading models. Argon can be applied to all stages of the robot's lifecycle, including mobile manipulation practice using MuJoCo, zero-shot robot control, and closed-loop steering of video world models.
Prime Intellect BlogPrime Intellect acquires Moondream, its founder will join Prime Intellect to oversee reasoning and cross-modal model development. Moondream 3.1 performs excellently in object detection, aerial images, charts, and crowded scenes (such as shelves), processing 34 requests per second, 24 times faster than the next accurate open-source model, and costs only one-thousandth of Meta's new model Muse Wonder when achieving 80% accuracy.
Amazon ScienceAmazon Science proposes Priming technology, which reduces the pre-training cost of hybrid state space models by over 99% by leveraging pre-trained Transformer weights, enabling developers to explore problems that were previously too costly to investigate.
Baseten Labs7575Base Labs research found that iterative learning with the talkie-1930-13b-it model using sports facts from 2011-2019 leads to fact forgetting, with forgotten facts still present in model weights. Experiments show that early learned facts are difficult to recall after 20 writes and merges, and are covered by later facts, but the covered facts can be relearned, primarily occurring in the later MLP weights.
The Verge · AI7575The Google AI Edge Foresight app can transcribe meetings and audio files offline, running on macOS and using the EmbeddingGemma 2 model. Users can simultaneously record handwritten key points and generate organized notes, with file, audio, and note data always remaining on the local device.
Google Developers BlogGoogle's AI Edge team has open-sourced ML Drift, a high-performance, cross-platform edge device GPU computing engine for on-device AI/ML inference. By abstracting the complexity of hardware low-level APIs such as OpenGL ES, OpenCL, Metal, and WebGPU, ML Drift enables developers to build real-time interactive AI experiences.
Amazon ScienceAmazon Science has released Graph Center Agent technology, which combines a continuous synchronized graph (digital twin) representing the network with device attributes and connections, with cascade graph algorithms (decomposition, aggregation, tracing) and agent orchestration to achieve minute-level root cause analysis of complex network failures, achieving 95% accuracy in a demonstration at NTT DOCOMO's MWC.
Google Developers BlogSparse VideoGen (SVG) dynamically routes attention heads to spatial or temporal masks by identifying the spatiotemporal structures of different attention heads in video diffusion models, converting theoretical sparsity into actual end-to-end speedup. On TPU v6e, SVG retains about 38.87% of query-key pairs, achieving 3.5x inference acceleration while maintaining output quality at a given sparsity level. The scheme employs custom JAX and Pallas Splash attention kernels to convert theoretical sparsity into physical sparsity on hardware.
Liquid AI Blog7575Liquid AI has released the DSpark Visual Language Model Accelerator LFM2.5-VL-3B, which boosts decoding throughput by 2.66x on GPUs and 3.13x on edge devices through a lightweight drafter model.
Liquid AI BlogLiquid AI announced at the Snapdragon Summit 2026 that its device-side context layer, Liquid Context, has been optimized for Snapdragon® processors, specifically using Qualcomm® Hexagon™ NPU.
Import AI7575Research shows that AI accelerates the discovery of network vulnerabilities but has a minimal impact on mathematical research, and AI research itself has not shown significant acceleration. The SPADE project generates training data in synthetic environments through self-play, with the Qwen3-30B model achieving the best results, significantly improving performance in both game and tool environment tasks. The Hawkeye tool is used to optimize GPU kernel construction.
Liquid AI BlogLiquid AI releases 4-bit checkpoints for LFM2.5-230M, LFM2.5-350M, LFM2.5-1.2B-Instruct, and LFM2.5-2.6B using Quantization-Aware Distillation (QAD) technology, restoring about 97% BF16 precision while maintaining Q4_0 GGUF low memory footprint and fast processing.
Liquid AI BlogFeatured7171Liquid AI releases LFM2.5-VL-3B, an upgraded 3.1B edge-side visual language model based on its predecessor LFM2-VL-3B, now available on Hugging Face and Playground
Why it matters: The original text provides 28 benchmark items, edge-side, and GPU throughput data, which can be used to compare with similar-sized competitors to determine the actual position of this 3B visual model.
Liquid AI BlogMacPaw and Liquid AI announce a strategic partnership to combine Liquid's base model with MacPaw's Elix and Mnemos technologies, creating the first Mac assistant Eney that integrates on-device intelligence, persistent memory, and native task execution, expected to launch later this year and expand to more products afterward
Liquid AI BlogFeatured7171Liquid AI has released LFM2.5-2.6B, which is an intelligent agent model that runs entirely on the device and can complete planning, tool calls, and multi-step tasks. Both the base and fine-tuned versions are now available on Hugging Face.
Why it matters: The original text provides the complete training process for the edge AI agent model and multi-benchmark comparisons, allowing you to judge the actual position of small models running AI agents locally.
Meta AI BlogMeta AI's DINOv3 and SAM models achieve real-time perception in edge computing, supporting the University of Pittsburgh's RAMMP project development of assistive robots, interacting through natural language and image data to reduce user cognitive load, already integrated in prototypes with automatic door button detection, cup recognition, and navigation assistance features, and plans to add voice and touch input support.
Berkeley AI ResearchAdapting to parallel reasoning allows models to decide on their own when to decompose and parallelize independent subtasks, how many threads to start, and how to coordinate. Existing methods such as fixed parallelism, heuristic structured search, and Monte Carlo tree search determine the parallel structure externally. Recent advancements like ParaThinker and GroupThink attempt to let models adaptively select parallel strategies, but decisions are still constrained externally.
Liquid AI BlogLiquid AI has released the LFM2.5-1.2B model family, with its pre-training extended to 28T tokens and significantly improved post-training performance through reinforcement learning, showing excellent performance on knowledge, instruction following, math, and tool use benchmarks. This series includes Base, Instruct, Japanese, Visual Language, and Audio Language models, all of which are open-source and available on Hugging Face and LEAP. The audio models are 8 times faster than their predecessors and natively run on vehicles, mobile devices, and IoT devices.
Cognition BlogSWE-1.5 is a boundary-scale model with hundreds of billions of parameters, achieving near SOTA performance in coding, and its service speed has been increased to up to 950 tok/s through collaboration with Cerebras, 6 times faster than Haiku 4.5 and 13 times faster than Sonnet 4.5. SWE-1.5 is now available on Windsurf.
Liquid AI BlogLiquid AI announces that its efficient Liquid Edge AI Platform (LEAP) platform now natively supports AMD's latest Ryzen™ and Ryzen AI™ processors, bringing performance improvements for developers building high-quality AI solutions on devices.
Dify BlogThe global consumer electronics leader serving over 100 million users has integrated generative AI into its operations through Dify.AI, bridging the gap between business and technology. Dify.AI's low-code platform allows novice users to quickly utilize mainstream models, expert users to independently customize solutions, and developers to enhance efficiency through component-based workflows. More than 200 AI applications have been created within a month, with 60 in use, reducing user voice analysis time from 8 hours to 3 hours.
