Hugging Face has introduced expert streaming for Mixture-of-Experts (MoE) models within the Transformers library. This technique allows running massive models that would typically exceed standard hardware memory limits by streaming active experts on demand.
PolyAI has introduced Dialog-RSN-1, a model specifically trained to handle live voice conversations by processing audio directly without intermediate transcription. It is now accessible in the AWS ecosystem for deployment as a private, secure inference endpoint.
Moonshot AI has made its 2.8-trillion-parameter Kimi K3 model accessible via API. The model supports a 1-million-token context window and is specifically notable for its performance in summarization and structured data extraction on long-form audio transcripts.
The DeepSeek V4.1 Flash model, featuring a 1M-token context window, is now available via the AssemblyAI LLM Gateway. It is particularly effective for cost-efficient batch summarization of hour-long audio recordings.
NVIDIA has released a new line of compact desktop PCs from ASUS, Dell, and MSI designed specifically for local AI workloads. These systems feature a 20-core NVIDIA Grace CPU, Blackwell RTX GPU, and up to 128GB of unified memory to support large language models and persistent agents.
Google has introduced Gemini 4 Neon, a new model designed for complex analytical, coding, and scientific workflows. It features a 1 million-token context window and emphasizes high steerability and efficiency for production-ready agentic tasks.
Researchers released Limite 1B - Violetto, a compact 1-billion-parameter model built for mathematical problem-solving. It supports a 131k token context window and is licensed under Apache 2.0.
Google has launched Genie 4, an advanced world model available to Google AI Ultra subscribers that enables users to create and edit interactive environments in real time. The update introduces native adaptive audio, 360-degree panorama grounding, and a new Project Genie agent for conversational environment creation.
Reson8 has launched Rumble-2-Flash, an open-source 0.5B parameter speech-to-text model supporting over 40 languages. It achieves performance parity with models up to 30B parameters on noisy, real-world call-center benchmarks.
Aionic Labs has launched OpenTSLM-Flamingo, a time-series language model (TSLM) now deployable via Amazon SageMaker JumpStart. It combines the Gemma 3 4B architecture with a time-series encoder to interpret operational metrics alongside text context.
AWS now provides managed support for NVIDIA Dynamo within SageMaker HyperPod, simplifying the deployment of large-scale language models. This update introduces multi-node inference, allowing teams to run models that exceed single-GPU memory through disaggregated prefill and decode worker pools.
Amazon SageMaker AI has launched a new 'KV_AWARE' routing strategy that directs LLM inference requests to instances based on their actual KV cache state rather than just prompt headers. This optimizes multi-turn conversations and long-context workloads by ensuring requests land where relevant data is already cached.
OpenAI has launched a public beta for the Responses API that allows you to select specific background processing windows. By opting for a two-hour or eight-hour completion window, your company can achieve substantial cost savings on token usage.
NVIDIA has released a technical primer on co-designing sparse attention for long-context models. By focusing on GPU-friendly indexer kernels—which identify relevant tokens before full attention—developers can significantly improve inference efficiency for context lengths surpassing 100K tokens.
NVIDIA unveiled 'Nemotron for Science', a toolkit combining the Nemotron 3.5 Super VL model with specialized chemistry datasets and post-training recipes. This platform allows researchers to deploy local agents that can autonomously plan experiments, execute code, and validate data findings.
Researchers introduced OmniTag-Bench, a new five-language framework for evaluating multimodal LLMs on simultaneous speech transcription and metadata tagging. Instead of using separate specialist models, the study demonstrates that a single, fine-tuned multimodal LLM can extract transcripts, emotion, gender, language, and named entities in a single inference pass.
Anthropic has implemented a text-based watermark for its newest Claude models, including Opus 4.6, Opus 4.5, Sonnet 4.5, and Haiku 4.5. The watermark applies when using default sampling configurations but may be absent if users modify temperature, top_p, or top_k settings.
OpenAI has introduced omni-moderation-csam-latest, a specialized model for identifying child sexual abuse material (CSAM) in images. This tool is available to approved organizations at no cost to assist in prioritizing content for human moderation.
Hugging Face has partnered with OpenRouter to power its language model inference. All language model requests now use OpenRouter to automatically select the optimal provider based on real-time price, speed, and reliability metrics.
NVIDIA has released the initial MLPerf Training v6.1 and Endpoints v1.0 results for its new Vera Rubin architecture. The system demonstrated significant leaps in both training speed and power efficiency for agent-based AI workloads.
Cohere has unveiled Video Intelligence within its Compass platform, allowing AI systems to perform granular, timestamped searches across video libraries. It processes both audio transcripts and on-screen visual information to retrieve precise moments that answer specific user queries.
Amazon has released a reference architecture that integrates the NVIDIA NeMo Switchyard library with LiteLLM to dynamically route coding agent requests between different models on Bedrock. By using stage or random routing, you can shift traffic between high-capability and high-efficiency models to balance performance and cost.
Amazon has added OpenFold3-preview to SageMaker JumpStart, allowing users to deploy models for predicting 3D structures of proteins, DNA, RNA, and ligands. The implementation utilizes asynchronous inference endpoints to handle long-running predictions while supporting cost-efficient autoscaling and scale-to-zero capabilities.
NVIDIA has released the cuda.bindings.cudla module, enabling direct Python-based management of the Deep Learning Accelerator (DLA) on Jetson AGX Orin. Developers can now orchestrate hybrid workflows where GPU preprocessing and DLA inference share a single CUDA stream.