Google AI Studio product and design lead Lei Xi uses the Gemini 4 Carbon model to develop a WindowsiOS project, running Windows 11 on Arm 24H2 system image on iPhone Duo via virtual machine, supporting touch operations. The project is open source, requiring Mac, development tools, and Windows ARM image deployment, can try to run some traditional Windows applications but compatibility is not guaranteed.
Baseten Labs6565Claude Code v2.1.159 released, adopting a gradient alignment reward mechanism, using the Qwen3-8B model on arithmetic tasks, rewarding by calculating the gradient similarity between student output and expert demonstrations, experiments show the model may overfit the precise wording of expert examples.
AWS Machine Learning Blog3434OpenFold3-preview can be deployed on Amazon SageMaker JumpStart, an open-source biomolecular structure prediction model developed by the OpenFold Consortium and AlQuraishi Lab at Columbia University, supporting 3D structure prediction for proteins, DNA, RNA, small molecule ligands, and their complexes.
AWS Machine Learning BlogDevelopers using coding agents like Claude Code and OpenAI Codex CLI need to be billed per task rather than per single call, so model routing should be based on the cost per task. NVIDIA NeMo Switchyard is an open-source library that helps AI agents select models to process each request, supporting Stage and Random routing strategies.
NVIDIA Technical Blog6565NVIDIA SOMA Retargeter can provide coordination and temporal reference for robot teaching through human action demonstrations, but it requires adaptation to different joint structures and activity ranges of robots. This tool can automatically adjust human actions to match robot characteristics, aiding in the development of humanoid robots.
NVIDIA Technical BlogTwo NVIDIA DGX Stations connected via ConnectX-8 RDMA, using vLLM and tensor parallel processing, successfully ran a 175B parameter GLM-5.3 NVFP4 model.
Ai2 Blog4040XYBench released, including 8K+ problem tests to evaluate how AI assistants handle requests based on misunderstandings; Claude Sonnet 4.6, GPT-5, Gemini 3.1 Pro, Qwen3.5-9B, and DeepSeek-V4 Pro often focus on the original incorrect method rather than the user's actual goal in the test.
MarkTechPostPerplexity releases open-source decision-making model Decider v1.3 and its lightweight version pplx-decider-flash-v1.3, adding a Locate feature to extract precise text segments.
Hugging Face Daily PapersOpen Vision Banana (OVB) has open-sourced its data, weights, code, and implementation details. Its 4B and 9B parameter-based open-weighted large model and 2.1M-3.8M training pairs surpass Vision Banana on Cityscapes with zero-shot performance of 73.9 vs 69.9 mIoU and can rival state-of-the-art affine-invariant depth estimators.
Ant Group inclusionAI · Hugging Face new modelsinclusionAI has released the Duo 1.2 causal-base model, a full-duplex multimodal world model for real-time audio-video interaction, featuring improved generation quality, and model weights can be downloaded from Hugging Face.
Ant Group inclusionAI · Hugging Face new modelsinclusionAI releases Duo-1.0-distill, a full-duplex multimodal world model for real-time audio-video interaction, including Realization weights with four-step distillation, which needs to be used in conjunction with the causal-base suite.
Ant Group inclusionAI · Hugging Face new modelsinclusionAI releases Duo 1.0 causal-base model, a full-duplex multimodal world model for real-time audio-video interaction, including implementation weights and backbones, intent models and adapters, audio-video encoders, speech recognizers, and forced aligners. The model weights are based on MiniMax H3 initialization and are open-sourced.
Ant Group inclusionAI · Hugging Face new modelsinclusionAI releases the Duo 1.0 Realization weights for FP8-quantized four-step distillation, used for its causal base model, supporting full-duplex real-time audio-video interaction, and needs to be overlaid with the causal base model weights.
Hugging Face Daily Papers6565Decision models, even as actors, can effectively teach reinforcement learning strategies by evaluating complete action trajectories, improving success rates on ALFWorld and WebShop.
Hugging Face Daily PapersDuo v1 introduces the Reciprocal World Model (RWM), which jointly models intentions and reality through autoregressive reasoning, synchronized attention, and dual inference distillation, achieving real-time full-duplex audio-video interaction. The model continuously perceives audio-video input while responding to users in real time, treating intelligence and physical reality as equal foundational components of the world, supporting interactions such as dialogue, sharing objects, and reshaping virtual scenes. The first batch of models has been open-sourced.
Hugging Face Daily PapersBirdDreams jointly generates bird's-eye view (BEV) videos and bird's-eye view videos, making scene layout a part of predicting the future. By providing an overhead view of vehicle position and movement through BEV, it reduces occlusion, while the perspective video retains scene details. On PhysicalAI-AV and nuScenes, BirdDreams more faithfully follows the input trajectory of its own vehicle and generates more reasonable self-motion without HD map or vehicle state input.
Ant Group inclusionAI · Hugging Face new modelsSingVL-4B GGUF is the GGUF version of the 4B parameter multimodal decision-making model Sing VL S1 based on Qwen3.5, which can answer named entity questions with text or JSON and up to four images, and uses a full forward pass scoring option, scoring 69.76 in the Text evaluation and 73.94 in the Vision evaluation. The model uses the llama.cpp adapter, with a default context size of 4096 tokens, supporting text and image input.
OpenRouter BlogWhen RAG applications exist correct information in documents but provide wrong answers, it is usually due to errors in the retrieval stage. Vector search can find relevant passages on the topic, but it often fails due to approximate matching, such as incorrectly calling the bucket limit of the Enterprise plan when asking about the Growth plan.
vLLM BlogDeepSeek-V4.1-Flash achieves a service capability of 1.7M tokens/s per GPU in a single NVIDIA Vera Rubin NVL72 rack, reaching up to 387 tokens/s per user, an increase of 58 times compared to GB300 NVL72.
vLLM BlogFeatured6060vLLM provides support for Inkling 2 Small on its first day of release. The model is a multimodal MoE with text and visual inputs, featuring a total of 594B parameters, 17B activation parameters, a maximum context length of 1M tokens, and comes with a built-in DSpark beam decoder.
Why it matters: The original text provides details of vLLM's Day-0 support for Inkling 2 Small and multiple kernel optimizations, allowing readers to understand the deployment path and throughput performance of the model.
Google Developers BlogGemini 4 Argon translates Go code panic paths into Z3 constraints, paired with Z3 to check trigger conditions, and paired with go test to replay trigger bytes, locating 3 embedded parser bugs and discovering 3 skill-lint potential bugs within 1 minute. This process is more efficient than go test -fuzz, capable of finding boundary check errors missed by fuzz testing.
Claude app release notesClaude Meetings begins Beta testing in Claude team and enterprise plans, allowing users to add Claude as a visible participant to Google Meet conferences, and to take notes, ask questions, and create content such as documents, whiteboards, and slides during the meeting.
Gemini 4 Hydrogen has been released, boosting the Artificial Analysis Intelligence Index score to 51 at the default high reasoning level (Argon is 50, with a maximum of 54), and typically outperforming similar models in task execution speed.
TechCrunch · AISame story · 3 outletsGoogle improves search complex question understanding ability through the newly released Gemini 4 Hydrogen model, which is 2.5 times faster and lower in cost than the previous generation. Hydrogen ranks 5th in Agent Arena for multi-step tasks, supports reasoning up to xhigh four tiers, and surpasses more powerful Gemini 4 Argon on one benchmark.
Google DeepMindFeaturedSame story · 3 outlets8787Google DeepMind releases Gemini 4 Hydrogen, targeting daily coding and intelligent agent workflows, with an Artificial Analysis intelligence index of 51 at the default high thinking level, higher than Argon's 50, and reaching 54 at the xhigh level
Why it matters: Google provides the division of labor and positioning of Hydrogen, Carbon, and Argon, as well as the actual differences in cost and speed between multi-model arrangements.
PyTorch Blog5050Helion analyzes kernel IR and data flow through compile-time heuristic algorithms, selecting fast settings without slow automatic tuning, classifying kernels as matrix multiplication, reduction, or point-wise operations, and choosing block size, warps, and phases to achieve near-automatic tuning speed.
Databricks BlogBlock company has launched Buzz, an open-source collaborative space, allowing employees to interact with AI agents in a shared workspace. Through Databricks Unity Gateway, users can access cutting-edge models from OpenAI, Anthropic, Google, and open-source models (specific number not mentioned in the original text), experimenting with different AI agent framework combinations. Employees can create AI agents and choose models (such as the lightweight version of Goose) to achieve multi-person collaborative development, analysis, and other tasks.
AWS Machine Learning Blog5858Amazon Bedrock Managed Knowledge Base achieves fine-grained access control for enterprise AI agents by injecting user context into the AgentCore Gateway tool layer and verifying JWT, combined with the ACL-aware functionality of the knowledge base.
AWS Machine Learning Blog5555AWS introduces Agent-to-Agent (A2A) protocol, allowing multi-agent systems operating across organizational boundaries to discover each other, communicate, and exchange tasks through a standardized protocol without exposing internal details or relying on the same framework.
IBM Research BlogIBM Research is studying two quantum computing hardware architectures, one of which is the spin qubit architecture that encodes information by controlling the spin state of electrons. Spin qubits use quantum dot structures (such as silicon-germanium heterostructures) to trap individual electrons, controlling their spin direction (up or down) with voltage pulses to represent 0 or 1. This architecture has high gate fidelity, scalability, relatively fast speeds, and simple control signals (requiring only voltage pulses), but the current number of qubits is less than that of superconducting qubits and faces challenges such as uniformity and manufacturing yield.
Galaxy Smart Learning has launched an AI education product, Xiao Lu Ai Xue, which adopts a dual-core architecture of the Doubao large model and a self-researched intelligent learning engine. It achieves learning situation diagnosis by analyzing students' answer records and develops functions such as learning situation diagnosis, ability assessment, learning difficulty matching, and task planning. The student-side intelligent agent recommends learning tasks and explains the basis, while the teacher-side intelligent agent summarizes student performance for personalized guidance reference. The product has served over 500 schools, covering 24 provinces, and supports teachers in adjusting teaching strategies through a human-machine co-education model.
PyTorch BlogThe EUROPA Union successfully trained a MoE model with 951.9B parameters on the JUPITER supercomputer using PyTorch and TorchTitan, increasing single GPU throughput from 90 to approximately 1,300 tokens/s with 384 NVIDIA GH200 GPUs.
OpenRouter BlogOpenRouter currently supports 8 languages including Simplified Chinese, Spanish, Japanese, and Korean, and localization features have entered Beta testing. Its system achieves self-correction by detecting changes in strings in the code daily, translating them using GPT-6.1 Sol, and then verifying through code checks and a second model. When translation fails, the system retries with the reason, recording errors as rules or test cases. The entire process runs within OpenRouter, with translation data stored in the code-side Git repository.
Latent SpaceSame story · 7 outletsTongxin Intelligence has completed tens of millions of yuan in seed and angel funding within half a year, with an average chip industry experience of over 20 years for the founding team. The company designs native chips supporting vector, tensor computation, and physical intelligence based on the CUV cognitive architecture proposed by Professor Zhu Songchun. The first product is positioned as a high-value inference chip, targeting Nvidia's Thor series on the edge and B300 series on the cloud, with FPGA prototype verification planned for early December. Funds will be used for core architecture design and R&D before chip fabrication.
AWS Machine Learning BlogJava teams can leverage Amazon Bedrock and Spring AI to build autonomous AI agents, addressing scalability, governance, security, and operational challenges. Spring AI brings AI capabilities to the JVM through the Spring framework (auto-configuration, dependency injection, composable advisors), while the Amazon Bedrock AgentCore platform supports framework and model-agnostic agent construction, connection, and optimization.
