AWS Machine Learning Blog2828Denodo platform builds a semantic layer on AWS, providing controlled enterprise data access for Amazon Bedrock AgentCore, supporting federated queries across Amazon Aurora, Amazon Redshift, and local PostgreSQL databases.
AWS Machine Learning Blog4040Eon replaced its local Model Context Protocol (MCP) server with Amazon Bedrock AgentCore Gateway, integrating cross-AWS, GCP, and Azure operational tool access into a single authentication gateway, reducing new engineers' investigation tool configuration time from 2 days to 2 minutes.
AWS Machine Learning Blog3131Canadian organizations can use OpenAI GPT-5.6, GPT-6, and GPT-6.1 models on Amazon Bedrock in the Canadian (Central) and Canadian West (Calgary) regions to achieve cross-region inference.
AWS Machine Learning Blog3838The Octave Reliance team built a three-tier secure multi-tenant AI infrastructure based on Cohere Embed v4 on Amazon Bedrock to solve the knowledge island problem in regulatory industry quality management, reducing deployment time by 50%.
AWS Machine Learning Blog4040GPT-OSS-20B is an open-source inference model that thinks in English by default, and only 22% met the target language requirements in multilingual inference tests. By using supervised fine-tuning (SFT) and reinforcement learning (GRPO) on Amazon SageMaker HyperPod, the team can improve the model's inference capabilities in French, German, Spanish, and Italian, where SFT is responsible for primary behavior teaching and GRPO optimizes when there is room for improvement.
AWS Machine Learning Blog3838Amazon Quick combined with Amazon Bedrock AgentCore provides supply chain teams with real-time data access across systems. Users can get cross-system answers in seconds through conversational queries, such as identifying high-income SKUs and costs affected by port closures. Quick serves as the interface layer, and AgentCore as the intelligence layer, connecting backend systems like ERP, monitoring, and procurement via the MCP protocol to achieve a complete process from problem discovery to generating emergency purchase requests, all operations requiring manual confirmation.
AWS Machine Learning Blog3939i14 deploys its 24-layer decoder Laplace 3B (3,072D parameters, using Gated Linear Attention architecture) to AWS Trainium2, achieving a training speed of 120,000 tokens per second and concurrent inference capability for 11,544 output tokens.
MIT Technology Review · AI5858Goodfire researcher Nick Cammarata published a blog post stating that LLMs might be easier to understand than previously thought, causing shockwaves in the AI world. Cammarata believes that the mechanisms of neural networks may not be as difficult to decipher as previously thought, a view that is particularly noteworthy against the backdrop of frequent LLM proxy out-of-control incidents.
Databricks BlogDatabricks has integrated the RFD4-Proteina model into Genesis Workbench, supporting execution in AI runtime environments and connecting research data and workflows. Developers can build and adjust pipelines using Genie Code and the NVIDIA BioNeMo Agent Toolkit, leveraging serverless NVIDIA GPU computing for protein design.
vLLM BlogAnt Group and the vLLM community open-source Humming, a high-performance quantized inference kernel library. Humming supports NVIDIA GPUs starting from the Turing architecture and CUDA-compatible devices like the T-Head Zhenwu 810E, providing extensive quantization support covering various integer and floating-point types, bit widths, and scaling schemes. It significantly outperforms existing implementations such as vLLM Marlin, FlashInfer, and DeepGEMM on multiple representative workloads.
Baseten Labs6060Baseten Labs releases the world's top decision models, Elephant-1 (9B parameters) and Mammoth-1 (27B parameters), both ranking first in the Jev Decision Index and Artificial Analysis Decision Index evaluations.
a16z News7070OpenAI has acquired Orchestra Bio. Current cutting-edge models like GPT-6 Astra, Fable 5.1, and Grok 4.6 can generate CAD, design PCBs, and describe assembly, but hardware design benchmarks like CAD Arena, EE Bench, and LuxoBench show that improvements are still needed. AI in hardware design faces the lossy problem of the interface between the digital and physical worlds, even if it can generate designs, a lot of work is still required.
Mistral AISame story · 2 outletsMistral AI has launched Voxtral 3, which includes an advanced voice model capable of real-time multi-step conversations in over 50 languages, along with a voice agent SDK. Its voice-to-text model, Voxtral Small Transcribe 3, achieved a Forced WER of 0.9% on the FLEURS 45 benchmark.
How the accounts differ:The Decoder mentions a text-to-speech model (Voxtral Mini TTS 3) that is not detailed in the first report.
Hugging Face Blog5858AnyLanguageModel 1.0 is released, supporting local model advancements on Apple platforms for a year. New support added for cloud models such as OpenAI Chat Completions, Anthropic Messages, and Google Gemini. Local model support includes llama.cpp, MLX, and Core ML. Tool calling and guided generation features are expanded. Cloud models have the same functionality as local models, and API keys support dynamic acquisition through closures.
Browser UseBrowser Harness achieves a score of 60.4 in matching evaluation using 2.1M median input tokens, matching Playwright MCP and Chrome DevTools MCP, becoming the browser control tool with the lowest cost per task at $0.042. Playwright MCP and Chrome DevTools MCP are faster but use more tokens.
Hugging Face BlogHugging Face has released TorchTPU, making TPU a native backend device for PyTorch. Transformers, Diffusers, and Kernels can now run directly on TPU. Users only need to select a model from the Hugging Face Hub and specify device_map="tpu".
Meta EngineeringGenRecs Tokenization enhances conversion rates by 0.35% for advertisers by representing items like ads and products with a stable semantic vocabulary, combined with Meta's Generative Ad Model (GEM) and adaptive ranking model. This technology leverages the world knowledge of pre-trained large models to enable recommendation systems to understand preferences of new items or users with no historical interaction, accelerates learning through shared concepts, achieves cross-activity transfer, and improves performance as large model scale increases, forming a feedback loop for performance improvement.
Meta EngineeringMeta engineers built a lean router with 1,300 lines to assign over 20 workflows to 14 specialized AI agents, with the router itself not performing any domain-specific tasks, significantly improving the efficiency and accuracy of data pipeline agents.
Sakana AI5555Sakana AI proposed 'QD-Rec', recommending candidate-hypothesis pairs as units, with generative AI writing new hypotheses based on past excellent hypotheses in a cycle to accumulate high-quality and diverse recommendations. Experiments were conducted in M&A, customer targeting, and paper recommendations, achieving high average evaluations in many settings and building large sets of recommendations.
LangChain BlogNu company, while developing AI health coach Nu, prioritized building evaluation criteria to avoid boundary-crossing medical diagnoses. Using the LangGraph framework and LangSmith tools, it ensures the system meets health guidance needs while complying with regulatory requirements through LLM judges scored by category and open-ended behavior tests. The evaluation framework is designed separately for single responses and complete conversations, supporting rapid migration between two leading model providers. Safety probes and guard nodes review conversation content in real-time, ensuring the product stays within compliance boundaries.
n8n Blogn8n workflows can run locally on NVIDIA DGX Spark without calling cloud APIs. By deploying the 35B hybrid expert model Qwen3.6-35B-A3B (NVFP4 format) and 8 other models on DGX Spark, tasks such as transcription, OCR, and image generation are supported. All models are accessed by n8n through OpenAI-compatible endpoints. Deployment requires caching models and using version 0.30.0 of vLLM to ensure offline operation and license compliance.
IBM Research Blog2222Coursera's new course 'AI Data Stream Architecture: From Principles to Practice' teaches engineers how to work across GPUs, TPUs, Cerebras WSE, Groq LPU, and IBM Spyre accelerators. The course is taught by Kaoutar El Maghraoui, a senior research scientist at IBM Research, covering execution models, compilation, memory systems, debugging, and performance modeling through six courses and hands-on experiments, emphasizing how different architectures affect model performance and the key role of compiler optimization.
Tencent Hunyuan · Hugging Face new modelsThe Hy-Memory-30B-A3B-GGUF model quantization file released by Tencent's Huayan uses a hybrid precision quantization scheme with imatrix calibration and MoE perception, achieving efficient local inference on llama.cpp, where the ~8 GB expert-mixed build file size is reduced to about 24.71 GB, with performance only 1.53% lower than Q6_K.
vLLM BlogvLLM achieves a downtime reduction of DeepSeek V4 Pro from 16GB200 to 20GB200 to 0.82 seconds and from 20GB200 back to 16GB200 to 1.96 seconds through asynchronous preparation and NIXL EP technology.
Inception Labs BlogMercury 2 decodes over 1000 tokens per second on standard NVIDIA GPUs, nearly 2 times faster than Gemini 3.1 Flash Lite in the query planning stage, 4.7 times faster than Claude Haiku 4.5, and 10 times faster than GPT-5 Mini. Each query step is faster than other models, allowing it to complete every step of the search process within the existing latency budget.
Hugging Face BlogTorchNeuron provides a new native backend for PyTorch on AWS Trainium, with native support for Transformers, Diffusers, and Kernels, making it unprecedentedly easy to run models on Trainium and share models on HF Hub.
MiniMax NewsMiniMax M3.1 Flash is officially launched, with nearly a threefold increase in online inference speed and a 70% reduction in cost compared to the previous generation. The model performs excellently in benchmark tests for complex intelligent tasks (cowork), computer use, software development, and network security, ranking highly in the AutomationBench-AA for automated tasks, scoring 73.7% on DeepSWE v1.1, 61.7 on OSWorld V2, 29.6 on ALE, and 82% on CyberGym.
Ollama BlogCloudflare's 27B multimodal decision model Clef has landed on the Ollama cloud platform, users can run it locally or use cloud services. Clef transforms state and typed question patterns into decisions, supports text and image input, is compatible with Jev and System One APIs, is fine-tuned based on Qwen3.8-27B, and adopts the Apache 2.0 license. In Cloudflare's decision index evaluation, Clef outperforms Jev and performs leading in most benchmark tests.
Inception Labs BlogInception Labs releases Mercury 3, this diffusion large language model surpasses GPT-6 Luna and Claude Haiku 5.5 in multiple benchmark tests, with a speed increase of 4-8 times reaching over 1,000+ tokens/second, expanding the context window to 1M tokens, with prices per million input/output tokens being $0.20 and $0.70 (50% discount at launch).
Inception Labs BlogMercury 2 is released on Azure AI Foundry, the world's fastest reasoning language model, built to make production AI feel instant. The model combines Inception's breakthrough diffusion architecture with Azure's enterprise-grade infrastructure, providing developers with models that offer both exceptional speed and quality.
Inception Labs BlogMercury 2 achieves a 78% task success rate on PinchBench, the fastest execution speed in its class, with a cost of $0.25 per million inputs and $0.75 per output, 4 times lower than Claude 4.5 Haiku, defining the feasible boundary for AI agents that balance performance, speed, and price.
MIT News · AI7575The multimodal AI model developed by MIT researchers combines 3D atomic structures with natural language descriptions, generating rich structural language descriptions from experimental research papers and first-principles calculation databases, helping AI learn the key properties of materials. The model outperforms text-only models in identifying solid-state lithium-ion battery electrolyte materials and can learn geometric structure representations, providing new research tools for materials science.
Google ResearchGoogle Research releases the VERID framework, enabling AI agents to run in physics-based, safe, and resource-efficient experiments; demonstrates closed-loop intelligent agent scientific discovery on adapter and de novo protein design prototypes; through three rounds of design-test-iterate cycles, AI-driven rational and de novo design will increase confirmed target hit rates to 33.5% and 42.1% on myoglobin and α-thrombin, respectively.
vLLM BlogvLLM has built a unified CuTe DSL decoding path for Kimi K3 through intelligent agent-assisted development, achieving a batch 1 performance of 840 tokens/s on 16 NVIDIA GB300 GPUs, with a 1.5-2x increase in decoding throughput in the test configuration, and a 1.8x increase in AgentX workload P90 interactivity.
Baseten BlogMeta releases Muse Spark Open, an open-source weight model with 816B parameters designed for code agents, featuring 42.7B active parameters, a 1M token context window, and full-modal input (text, image, video, audio), performance comparable to cutting-edge lab models, using a compressed MoE architecture and hybrid global/local attention mechanisms to optimize inference efficiency, reducing model size to 540GB via NVFP4 checkpoints.
Microsoft AI NewsMicrosoft AI has completed the pre-training of a new MoE large model MAI-Base-2, with 80 billion active parameters and 2.8 trillion total parameters, surpassing all open-source models in internal evaluations. The model is 4-12 times more efficient in pre-training FLOPs than the strongest open-source base model DeepSeek-V4-Pro-Base and is used to train larger models. Microsoft plans to license the model weights through Foundry using organic data pre-training and full-stack control.
