NVIDIA Technical Blog6565The NVIDIA Alpamayo 2 Edge edge inference model can run within automotive computing and latency constraints. The model achieves 70.5 on the MMLU benchmark, supports an 8k context window, speeds up 3.5x, and costs $3/ million tokens with closed-source deployment.
MarkTechPostZyphra has released ZONOS3-Mini, an open-source speech synthesis model with 800M total parameters and 100M active parameters, supporting real-time local deployment and zero-shot voice cloning. The model ranks first in the open model category in Zyphra's self-developed ZTTS2-Eval benchmark test (66.2 points), with the first audio generation on a single H100 GPU taking 80 milliseconds, but the reference audio cloning effect is far lower than the embedded mode.
Mastra BlogMastra FDE is a pre-deployment engineering plan for all teams building AI agents. By deploying engineers familiar with Mastra at a cost lower than hiring one senior AI engineer, it assists teams from the first week in architecture design to production deployment, and provides regular 1:1 communication, consulting, or proactive development support. It also includes access to a Slack channel with founders and engineers, quick response, roadmap insights, customized training, and the full suite of Mastra Enterprise features.
OpenAI NewsA new study by OpenAI and Upwork shows that the demand for freelancers is changing, requiring a new set of skills, and the budget and duration of AI-related projects are increasing.
Amazon ScienceMitra-v2 leads the single-model field in the TabArena benchmark, outperforming Mitra-v1 on 33 datasets and ranking first on ScoringBench.
NVIDIA Technical BlogNVIDIA has released the open-source NVKE and NVKR projects for Kubernetes management services for GPU architectures, both in public preview. NVKE and NVKR are designed to help AI factory customers consume GPU resources through Kubernetes APIs, tools, and ecosystems.
NVIDIA Technical BlogNVIDIA Snapshot enables GPU workloads initialized to be reused across inference replicas, accelerating startup and recovery. When inference demand surges or replicas fail, available GPUs do not immediately translate into service capabilities. Initializing LLM workloads (cold start) may take several minutes.
NVIDIA Technical Blog6565Agent Substrate addresses the inefficiency of Kubernetes when running AI agents by sharing Pod resources to reduce idle waste and maintain agent states, shortening cold start time by 3.5 times compared to traditional methods. The framework has been donated to the Linux Foundation to support large-scale agent deployment.
MiniMax H3-Image released and weights opened, capable of generating 4096x4096 pixel 4K images in 10 seconds. The model has text-to-image generation, image reference, and editing capabilities, achieving speeds five times faster than mainstream market models, supporting 16 reference images, and offering a 50% discount on 16MP generation requests in the first week.
Ars Technica · AISame story · 3 outletsDeepMind releases Gemini Robotics 2.5, whose robotic AI has approached human level in completing 200 tasks such as gripping food and laboratory pipetting, and has achieved high-precision operations like tying bows on the Apptronik Apollo humanoid robot; this version significantly improves dexterity while launching the open-source security benchmark "ASIMOV-jailbreak".
The Verge · AISame story · 2 outletsAdobe announced that it will introduce chatbots and AI agents as the next-generation design assistants, even automating creative workflows and repetitive tasks. At the Adobe Max conference in Miami, Adobe introduced a feature that allows third-party AI agents to access and modify Creative Cloud projects, new features for the Firefly AI assistant, and the ability of AI agents to perform repetitive tasks in the background.
How the accounts differ:The Verge states Adobe will automate creative workflows and repetitive tasks; TechCrunch only mentions injecting AI features into products.
TechCrunch · AIKora Systems released the Linux desktop operating system named Kora on November 10, 2026, integrating AI models and agents, supporting cross-application collaboration, file access, and context preservation between tasks, planned to be launched for free in 2027, accelerated development based on OpenAI Codex and Anthropic Claude, targeting heavy AI users.
TechCrunch · AISame story · 2 outletsAdobe announced at its MAX conference that it will add more AI features to Photoshop, Premiere, Illustrator, and InDesign, while expanding the AI capabilities of Firefly and launching a creative AI agent.
How the accounts differ:The Verge states Adobe will automate creative workflows and repetitive tasks; TechCrunch only mentions injecting AI features into products.
Hugging Face Daily PapersFlide is an open-source reasoning engine that enables multiple problems to share computation through precise state branching technology, supporting over 30 open-source decision models on Decision Index 0.4.1, with inference speed 9.5 times faster than vLLM and SGLang.
n8n Blogn8n Labs has released an independent personal intelligent agent called Nod, an open-source Agent that runs on a user's local computer. Nod has its own computer and browser and can complete tasks that chat tools cannot handle, such as unsubscribing, monitoring prices, booking hotels, etc., and can execute multiple tasks in the background in parallel. It supports users in selecting AI providers and has long-term memory capabilities, scoring 93.2% on the LongMemEval benchmark test.
Hugging Face Daily PapersFrontier3DBench uniformly evaluated the ability of code agents in object, scene, and interactive 3D world creation across 145 tasks, GPT-6.1-sol performed best at the object and scene levels, with a total score of 66.02/100, but all configurations had reference reconstruction scores lower than text-driven generation in object and scene tasks. The generated world averaged 83.08/100 in local mechanisms, but only 20.82 and 30.26 respectively in motion removal and stateful workflow, showing that coordinated activity completion remains a core bottleneck
Blender 5.3 has been officially released, natively supporting 3D Gaussian splatter (importing PLY/SPZ format data), switching the default graphics backend for the Linux platform to Vulkan (minimum version 1.1), and synchronously adjusting the Windows version except for ARM devices.
Hugging Face Daily Papers6565Reframe enhances the tool use performance of embodied visual agents through evidence-aware reinforcement learning and observation gatekeeping strategy distillation, outperforming Mix-RL and basic MOPD methods in 14 image and video benchmarks. Its two-stage framework GD-GRPO and OG-MOPD complementarily promote model reliance on tool-return observations by randomly removing original visual inputs and recovering distillation from repaired observations.
Google DeepMind5858Google DeepMind released two foundational datasets, SatWeave and PARDIG, to support the development of weather, hydrological, and climate applications.
A user developed a browser-based tool that allows inputting a location to query all satellites passing over it in the past 7 days and their real images (Sentinel-2/Landsat/NASA), rendering approximately 17,000 on-orbit objects via GPU and calculating Kepler's equations in real-time to achieve high-precision positioning. The tool only displays real satellite data and does not store user locations.
The SpaceX SI team completely rewrote Grok reinforcement learning and reasoning software from scratch, primarily developed in C with a small amount of C++, replacing JAX and SGLang. This underlying engineering accumulation is the core foundation of Grok's new AI stack, reusing Starlink's engineering expertise to support Grok's iteration.
openJiuwen releases the TTSE (Dual-Track Self-Evolution) framework, dividing Agent experience into FACT (Environmental Facts) and TIP (Conditional Techniques) managed on dual tracks, continuously calibrating environmental cognition and execution methods through real task feedback.
Hugging Face Daily PapersReproBridge enhances the success rate of machine learning experiment replication through standardized design (refined codebase, environment setup, experiment launchers, and replication guides), increasing the average replication rate of four models from 76.4% to 92.6% on ReproBridge-Bench. Refined artifacts reduce the average downstream API cost by 19.7%, narrow the success rate gap between models to 3.9 percentage points, and achieve a success rate above 90% for all models. Removing code or environment components significantly reduces replication effectiveness.
Hugging Face Daily PapersLingBot-Touch is a tactile foundational model that learns a unified, scalable representation across different tactile sensors (resistive, magnetic, visual tactile), producing plug-and-play tactile tokens, and continuously enhancing the dexterity of diffusion strategies and visual language action (VLA) models (such as π_{0.5} and LingBot-VLA). Experiments show that this tactile representation significantly improves performance in complex manipulation tasks in both simulation and real robot deployment while maintaining strong computational efficiency.
Hugging Face Daily PapersA precise tile attention algorithm integrates additive relative position terms into the computation similar to FlashAttention, reducing memory and computational overhead, accelerating long-context relative position Transformers. The method (1) uses offset-aware tiles to avoid redundant index and memory access, (2) only computes necessary key/query relation products and reuses them in forward and backward passes, and (3) uses offset-based gradient sharding and aggregation to eliminate atomic contention.
Hugging Face Daily PapersOneTrans-GSR generates unified recommendation ranking lists through a single autoregressive Transformer, improving business metrics and reducing re-ranking and hybrid sorting latency. After deployment in industrial recommendation systems, GMV/u increased by 2.63%, advertiser value increased by 2.06%, and repurchase rate increased by 2.03%, while end-to-end latency was reduced by 13%.
Microsoft has announced a new patent that allows users to contribute idle GPU computing power to participate in AI training and get paid. The system breaks down training tasks into micro-tasks and distributes them to idle devices for execution, with users able to earn points to exchange for free software, cash, etc. Microsoft plans to open this solution to third-party organizations to alleviate the issue of AI training computing power shortages, but users need to actively authorize and privacy protection issues remain to be resolved.
Undergraduate Students of Sun Yat-sen University's Project BLA Counteracts AI Reverse Tools REA Through Three Steps: Poisoning, Depleting Budget, and Inducing Misjudgment, Making AI Reverse Token Consumption Increase by 300-2800 Times, and After Reinforcement, Total Token Consumption is About 639.5 Times That Before Reinforcement, and Total Cost is About 304.4 Times. This Project Has Received 3900+ Stars and 476 Forks.
e-means achieved a test loss lower than Lloyd's algorithm within 25 rounds of training, requiring only 1.2 to 3 rounds. Compared to the state-of-the-art Flash-KMeans, GPU end-to-end speed improved 10-45 times, and CPU end-to-end speed improved 8-15 times. e-means achieved dual optimization of speed and loss through exponential moving average updates, automatic hyperparameter adjustment, improved initialization, and dead cluster revival strategies.
How the accounts differ:Hacker News states e-means achieves lower test loss in 25 rounds, while MarkTechPost mentions 1.4 to 2.6 rounds to match k-means 25 rounds' loss level.
Claude Fable 5.2 can generate a complete game from a set of paintings by a nephew with one click, including 175 images and 140 audio clips. The generated action game features 18 characters, 12 levels, and a storyline, supports 4-player online play, and is achieved through Claude Code and Opus 4.8.
Hugging Face Daily PapersWorldFoundry provides open-source infrastructure for reproducibly reasoning and evaluating generative world models across video, 3D/4D, embodied action, and interactive systems, including traceable evidence and benchmarks, featuring 270 model entries, 74 benchmarks, and 62 registered metrics.
Hugging Face Daily PapersThe A2-MEAS framework enhances robot control and fault recovery by learning correction codes and embodied visual language action models, achieving a 0.5755 Q-score on the 50-task BEHAVIOR 2026 public subset and a 45.40% task success rate, with an average success rate of 57.92% on 12 real robot tasks. The framework employs action pre-training and task specialization, with a training throughput of 427.1 samples/second, and the system will open-source the model, training scripts, and infrastructure.
Claude Code releases1515The Claude Code v2.1.322 version has added a fast mode in the global, US, and European regions on the Vertex platform. If Vertex rejects this mode, it will fall back to standard speed.
vLLM releases1515vLLM v0.34.0rc0 released, migrating Gemma and StableLM to the Transformers backend.
