{"digest":{"id":2,"generated_at":"2026-10-28T02:16:56.330Z","window_start":"2026-10-21T02:15:51.302Z","window_end":"2026-10-28T02:15:51.302Z","item_count":9,"headline":"AWS bets on serverless training, OpenAI, and Trainium maturity","narrative":"AWS shipped its most aggressive serverless model customization push yet. SageMaker AI launched training session APIs that let developers write custom training loops—including novel RL algorithms and loss functions—via stateful API calls while AWS manages the GPU fleet and bills per token processed. This isn't managed notebooks with prettier UIs; it's infrastructure-free training for practitioners who want control without cluster management. Paired with custom sequence lengths up to 16K tokens and support for Muse Glimmer, Qwen 3.8, and Ministral 3, AWS is signaling that post-deployment model evolution belongs in SageMaker, not just Bedrock.\n\nOn the infrastructure layer, AWS Neuron 2.33 shipped TorchNeuron in public beta alongside Neuron Observer for sub-second Trainium health detection and vLLM 0.29.0 support. The 25 new experimental NKI kernels for frontier models like DeepSeek-V3.2 make clear that Trainium is no longer a cost-play footnote; AWS is treating it as a first-class platform for training and serving. The Reflection Beam model launching on both SageMaker JumpStart and Trainium3 reinforces the pitch: if you want an open-weight frontier model you can inspect, fine-tune, and deploy entirely in your account, AWS has the silicon and software stack ready.\n\nBedrock's role as the neutral frontier-model marketplace also advanced. OpenAI's GPT-6.1 Terra arrived with multimodal reasoning, computer use, and a 1-million-token context window at roughly half the cost of GPT-6.1 Sol. AWS isn't window-shopping OpenAI models after Azure; it's getting them day-and-date and pricing them aggressively. That, plus the SageMaker serverless training APIs, gives builders two clear paths in one cloud: consume frontier models through Bedrock, or own and customize open models on Trainium without touching a cluster.\n\nWhat builders should notice is the narrowing gap between \"easy\" and \"expert.\" AWS wants advanced practitioners to use serverless training APIs for custom RL loops while their colleagues call GPT-6.1 Terra through Bedrock. The platform bet is that both workflows run on the same account, the same IAM, and increasingly, the same billing model per token. If you're choosing between clouds, AWS is arguing that you don't need to pick between managed convenience and infrastructure control—you can switch between them without switching providers.","themes":[{"name":"Serverless model customization","summary":"AWS is aggressively abstracting GPU infrastructure for training with stateful APIs for custom loops, per-token billing, and expanded model support."},{"name":"Trainium ecosystem maturation","summary":"Native PyTorch support, vLLM integration, and frontier-model kernels show AWS silicon becoming a genuine alternative to NVIDIA."},{"name":"OpenAI partnership deepening","summary":"Rapid Bedrock availability of GPT-6.1 Terra at half the cost of Sol positions AWS as the neutral frontier-model marketplace."}],"model":"cloudflare/@cf/moonshotai/kimi-k2.6"},"lastRun":{"id":2,"started_at":"2026-10-28T02:09:55.354Z","finished_at":"2026-10-28T02:16:56.429Z","fetched":100,"new_items":0,"status":"ok","error":null},"items":[{"id":"whats-new-v2#sagemaker-mlflow-apps-now-support-mlflow-3-16","title":"SageMaker MLflow Apps now support MLflow 3.16","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/mflow-version-3-16/","posted_at":"2026-10-27T20:40:00Z","summary":"SageMaker MLflow Apps upgraded to MLflow 3.16, adding a redesigned trace explorer with customizable columns and faster metric charts for generative AI debugging and experiment tracking.","theme":"sagemaker-ml","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-sagemaker"]},{"id":"whats-new-v2#gpt6-1terra","title":"OpenAI GPT-6.1 Terra is now generally available on Amazon Bedrock","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/openai-gpt-6-1-terra-on-amazon-bedrock/","posted_at":"2026-10-26T18:27:00Z","summary":"OpenAI's GPT-6.1 Terra is generally available on Amazon Bedrock, offering multimodal reasoning, computer use, and a 1-million-token context window at roughly half the cost of GPT-6.1 Sol, signaling AWS's continued deepening of its OpenAI partnership.","theme":"models","significance":3,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#launch-dotnwvbapafnhtgielb5qj3","title":"Amazon SageMaker AI announces training session APIs for serverless model customization","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/training-session-apis-on-sagemaker-ai","posted_at":"2026-10-22T16:30:00Z","summary":"SageMaker AI launched training session APIs for serverless customization, letting developers write custom training loops—including novel RL algorithms and loss functions—via stateful API calls while AWS manages GPU infrastructure, billing per token processed.","theme":"sagemaker-ml","significance":3,"fetched_at":"2026-10-28T01:58:04.998Z","services":["aiml","amazon-sagemaker"]},{"id":"whats-new-v2#6efe1ae4-af3a-40e7-b939-287645681c4d","title":"Amazon SageMaker AI now supports serverless model customization for Muse Glimmer, Qwen 3.8 and Ministral 3 models","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/amazon-sagemaker-muse-qwen-ministral/","posted_at":"2026-10-22T08:00:00Z","summary":"SageMaker AI expanded serverless model customization to support Muse Glimmer, Qwen 3.8, and Ministral 3 models using SFT, DPO, RFT, LoRA, or full fine-tuning.","theme":"sagemaker-ml","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-sagemaker"]},{"id":"whats-new-v2#e6c820f4-fd53-45d4-8757-de108252fe56","title":"AWS Lambda MicroVMs is now available in 4 additional AWS regions","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/lambda-microvms-4-additional-regions","posted_at":"2026-10-22T07:06:00Z","summary":"AWS Lambda MicroVMs expanded to four additional Asia Pacific and Europe regions, bringing its isolated, stateful serverless compute primitive for AI coding assistants and agents to 14 regions total.","theme":"infrastructure","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["aws-lambda"]},{"id":"whats-new-v2#6d3c719c-b0fa-4125-81ec-9d58cca3ded7","title":"AWS Transform adds storage migration assessments for Amazon S3, EBS, EFS, and FSx","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/transform-assess-s3-ebs-efs-fsx/","posted_at":"2026-10-21T19:10:00Z","summary":"AWS Transform added automated storage migration assessments that map on-premises NAS, SAN, and backup inventories to optimal AWS storage services with TCO projections.","theme":"other","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-fsx-netapp-ontap","aws-transform ","amazon-efs","amazon-elastic-block-store","amazon-s3"]},{"id":"whats-new-v2#beam-model-by-reflection-is-now-available-on-amazon-sagemaker-jumpstart-and-aws-trainium","title":"Beam model by Reflection is now available on Amazon SageMaker JumpStart and AWS Trainium","url":"https://aws.amazon.com/about-aws/whats-new/2026/01/reflection-beam-on-sagemaker-jumpstart-and-aws-trainium/","posted_at":"2026-10-21T15:00:00Z","summary":"Reflection's open-weight Beam model is now available in SageMaker JumpStart and optimized for AWS Trainium3, giving customers a controllable frontier model they can fine-tune, inspect, and deploy entirely within their AWS accounts.","theme":"models","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-sagemaker-jumpstart","aiml","amazon-sagemaker-deploy"]},{"id":"whats-new-v2#5dd94464-e4c2-4d43-8cf5-95bc9260dc85","title":"AWS Neuron 2.33 adds native PyTorch support, Neuron Observer, and other new features","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/aws-announce-neuron-2-33-0","posted_at":"2026-10-21T15:00:00Z","summary":"AWS Neuron 2.33 shipped TorchNeuron in public beta alongside Neuron Observer for sub-second Trainium health detection, vLLM 0.29.0 support, and 25 new experimental NKI kernels for models like DeepSeek-V3.2.","theme":"infrastructure","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":[]},{"id":"whats-new-v2#6386d2b7-5f71-4656-aec8-21c9a3fd89ec","title":"Amazon SageMaker AI now supports custom sequence lengths for serverless model customization jobs","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/sagemaker-custom-sequence-lengths/","posted_at":"2026-10-21T08:00:00Z","summary":"SageMaker AI serverless model customization now lets users configure dataset sequence lengths up to published per-model maximums (up to 16K tokens), enabling better fine-tuning for both short and long-context datasets.","theme":"sagemaker-ml","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-sagemaker"]},{"id":"whats-new-v2#launch-4os66lelsia7d84fwwutzs","title":"AWS announces TorchNeuron: Native PyTorch Support for AWS Trainium (Public Beta)","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/aws-announce-native-pytorch-support-trainium","posted_at":"2026-10-20T15:00:00Z","summary":"AWS launched TorchNeuron in public beta, an open-source native PyTorch backend for Trainium2 and Trainium3 that supports eager execution and torch.compile via a simple .to('neuron') device switch.","theme":"infrastructure","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-ec2"]},{"id":"whats-new-v2#84a56f02-d982-4510-909a-342c85bae148","title":"Structured outputs for Claude models available on Amazon Bedrock","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/claude-structured-outputs-bedrock/","posted_at":"2026-10-20T14:00:00Z","summary":"Amazon Bedrock now supports structured JSON outputs for all Anthropic Claude models via both bedrock-runtime and bedrock-mantle endpoints, removing the need for custom parsing to get schema-compliant responses.","theme":"bedrock-platform","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#170504bd-0a04-4be4-819c-4bb8c1780365","title":"Amazon Bedrock Guardrails now supports Standard tier for contextual grounding checks","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/bedrock-guardrails-grounding-standard/","posted_at":"2026-10-20T08:00:00Z","summary":"Amazon Bedrock Guardrails introduced a Standard tier for contextual grounding checks that better handles reasoning and conversational responses while improving consistency across differently worded answers, matching the Standard tier for content filters.","theme":"bedrock-platform","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#560d2070-50c9-434d-b234-80d45f7dc2f5","title":"Amazon Bedrock Advanced Prompt Optimization now supports prompt shortening","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/bedrock-prompt-optimization-shortening/","posted_at":"2026-10-20T08:00:00Z","summary":"Amazon Bedrock Advanced Prompt Optimization added an optional prompt-shortening mode that returns compressed prompts only when they maintain or improve evaluation scores, cutting input tokens and latency while providing visible compression ratios.","theme":"bedrock-platform","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#nemotron-3-5-super-vl-model-now-available-on-amazon-sagemaker-jumpstart","title":"Nemotron 3.5 Super VL model now available on Amazon SageMaker JumpStart","url":"https://aws.amazon.com/about-aws/whats-new/2026/01/nvidia-nemotron-3.5-super-vl-on-sagemaker-jumpstart/","posted_at":"2026-10-16T17:59:00Z","summary":"NVIDIA's Nemotron 3.5 Super VL vision-language model is now available on SageMaker JumpStart, offering a 120B MoE backbone with 1M-token context, multimodal agentic capabilities, and both NVFP4 and BF16 deployment variants.","theme":"models","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["aiml","amazon-sagemaker-jumpstart","amazon-sagemaker-deploy"]},{"id":"whats-new-v2#amazon-sagemaker-studio-makes-it-easier-to-get-started-for-new-ai-ml-builders-with-social-sign-in","title":"Amazon SageMaker Studio simplifies getting started for new AI/ML builders with new sign-in options","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/sagemaker-studio-new-sign-in-options/","posted_at":"2026-10-15T20:10:00Z","summary":"Amazon SageMaker Studio now lets new users sign up with Google, GitHub, or Apple credentials and start building in a pre-configured environment without traditional AWS account setup or credit cards, lowering friction for new AI/ML builders.","theme":"sagemaker-ml","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-sagemaker-studio","amazon-sagemaker-jumpstart","amazon-sagemaker","aiml"]},{"id":"whats-new-v2#43c2b47d-9cac-46a3-9a9a-59aa74e2a1c1","title":"AWS Transform for migrations is now available in Asia Pacific (Singapore)","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/aws-transform-singapore","posted_at":"2026-10-15T19:24:00Z","summary":"AWS Transform for migrations is now available in the Asia Pacific (Singapore) Region, letting customers run agentic AI-powered migration workflows with local data residency.","theme":"other","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["aws-transform "]},{"id":"whats-new-v2#muse-spark-open-model-now-available-on-amazon-sagemaker-jumpstart","title":"Muse Spark Open model now available on Amazon SageMaker JumpStart","url":"https://aws.amazon.com/about-aws/whats-new/2026/01/muse-spark-open-on-sagemaker-jumpstart/","posted_at":"2026-10-15T18:00:00Z","summary":"Meta's Muse Spark Open, an 816B-parameter MoE coding model with a 1M-token context window and multimodal input support, is now available on SageMaker JumpStart in both BF16 and NVFP4 quantized variants for agentic software engineering workflows.","theme":"models","significance":3,"fetched_at":"2026-10-28T01:58:04.998Z","services":["aiml","amazon-sagemaker-jumpstart","amazon-sagemaker-deploy"]},{"id":"whats-new-v2#6cad0ccd-0b6e-4e8c-8da6-c1be1d608eb4","title":"AWS HealthOmics now supports shared VPCs","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/omics-workflows-shared-vpcs/","posted_at":"2026-10-15T17:42:00Z","summary":"AWS HealthOmics now supports sharing VPCs across accounts via AWS RAM, letting organizations centralize network management while allowing individual accounts to run bioinformatics workflows in shared subnets.","theme":"other","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-omics"]},{"id":"whats-new-v2#a78ffde4-1b92-4dc2-aa46-b62084e3ea3f","title":"AWS HealthOmics expands container registry support for private workflows","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/aws-healthomics-expanded-container-registry-support/","posted_at":"2026-10-14T21:55:00Z","summary":"AWS HealthOmics expanded third-party container registry support to include Seqera Wave, NVIDIA NGC, and Google Artifact Registry, automatically copying images into Amazon ECR for private workflows without manual migration.","theme":"other","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":[]},{"id":"whats-new-v2#4605ab38-bb58-42ab-ac7c-15de3fbccf8d","title":"AWS HealthOmics introduces resource fallback order for Nextflow workflows","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/aws-healthomics-resource-fallback-for-Nextflow/","posted_at":"2026-10-14T16:00:00Z","summary":"AWS HealthOmics introduced resource fallback directives for Nextflow workflows, letting researchers specify ordered lists of preferred accelerator types—including CPU fallback—to automatically handle accelerator unavailability.","theme":"other","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":[]},{"id":"whats-new-v2#9bd791e0-41c4-4ded-9b63-5d90cd6f09f7","title":"Announcing consumption-based pricing for short-term memory in Amazon Bedrock AgentCore","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/agentcore-memory-consumption-based-pricing","posted_at":"2026-10-12T16:00:00Z","summary":"Amazon Bedrock AgentCore short-term memory switched from flat-rate per-event pricing to consumption-based billing on ingestion, retrieval, and storage dimensions, lowering costs for small-payload agent workloads.","theme":"agents","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":[]},{"id":"whats-new-v2#launch-psjcdtbzsc16pwbb5wsfzg","title":"Amazon Quick now supports brand templates for on-brand presentations and documents","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/amazon-quick-brand-templates-on-brand-presentations-documents/","posted_at":"2026-10-09T15:00:00Z","summary":"Amazon Quick now supports corporate brand templates extracted from PowerPoint and Word files, ensuring AI-generated presentations and documents adhere to approved colors, fonts, logos, and layouts.","theme":"applied-ai","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-quick-suite","aiml"]},{"id":"whats-new-v2#846734d0-1918-4072-b2c8-4485be430c34","title":"Claude Haiku 5.5 is now available on AWS","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/claude-haiku-5-5-aws/","posted_at":"2026-10-07T14:00:00Z","summary":"Anthropic's Claude Haiku 5.5 is now available on Amazon Bedrock and Claude Platform on AWS, offering subagent-optimized speed, effort controls, and a roughly 75% cost reduction over Haiku 4.5 for high-volume tasks.","theme":"models","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#launches-89218204-ed17-489d-a4b3-cdde0f8d49c6","title":"GLM 5.3 by Z.ai is now generally available on Amazon Bedrock","url":"https://aws.amazon.com/about-aws/whats-new/2026/10/amazon-bedrock-glm-5-3/","posted_at":"2026-10-05T21:42:00Z","summary":"Z.ai's GLM 5.3, a 753B-parameter MoE model with a 1M-token context window and up to 128K output tokens, is now generally available on Amazon Bedrock with explicit prompt caching support for agentic coding workloads.","theme":"models","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#gpt6uf","title":"OpenAI GPT-6 Astra now supports UltraFast mode on Amazon Bedrock","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/openai-gpt-6-astra-ultrafast-on-amazon-bedrock/","posted_at":"2026-09-30T19:00:00Z","summary":"OpenAI's GPT-6 Astra now supports an UltraFast mode on Amazon Bedrock, delivering up to 6x faster inference and 300 tokens per second for latency-sensitive applications like real-time coding assistants and interactive agents.","theme":"bedrock-platform","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#558ed2a0-2cfc-4d8f-8634-740d6d220ace","title":"Amazon Bedrock expands Claude models in-region support in the UK (London)","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/claude-region-expansion-lhr/","posted_at":"2026-09-30T17:00:00Z","summary":"Anthropic's Claude Opus 5.5 and Claude Sonnet 5 are now available with in-region inference on Amazon Bedrock in the UK (London) Region for data residency-sensitive customers.","theme":"bedrock-platform","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#launch-nalrspjaca1bkxa5txbspg","title":"Amazon Quick adds Hierarchy Filter for guided dashboard drill-down","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/quick-hierarchy-filter","posted_at":"2026-09-30T15:00:00Z","summary":"Amazon Quick added a hierarchy filter control that lets dashboard readers drill through up to five related dimension levels in a single dropdown, reducing clutter and guiding data exploration.","theme":"applied-ai","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-quick-suite"]},{"id":"whats-new-v2#be84f691-fb77-4dcb-9a84-21bc87279015","title":"Amazon Quick now supports live data from your datasets in apps","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/live-data-in-apps","posted_at":"2026-09-30T15:00:00Z","summary":"Amazon Quick now lets users build applications that pull live data directly from Quick datasets, enabling real-time operational workflows where row-level and column-level security is applied per user.","theme":"applied-ai","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-quick-suite"]},{"id":"whats-new-v2#bmapreview","title":"Amazon Bedrock Managed Agents, powered by OpenAI, is now available in preview","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/bedrock-managed-agents-preview/","posted_at":"2026-09-29T21:10:00Z","summary":"Amazon Bedrock Managed Agents, a joint AWS-OpenAI agent framework built on a customized OpenAI Agents API, is now in preview, letting developers build OpenAI-optimized agents with durable sessions, IAM roles, human approval, and MCP tool integration entirely inside AWS.","theme":"agents","significance":3,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#gpt-6-1-sol","title":"OpenAI GPT-6.1 Sol is now generally available on Amazon Bedrock","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/openai-gpt-6-1-sol-on-amazon-bedrock/","posted_at":"2026-09-29T20:17:00Z","summary":"OpenAI's GPT-6.1 Sol is now generally available on Amazon Bedrock, offering agentic coding and computer-use performance near GPT-6 Astra at roughly one-fifth the cost, plus explicit prompt caching for context-reuse workloads.","theme":"models","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#0a44fd84-1aa9-44f4-9e4d-810480f677fa","title":"Amazon Bedrock expands Claude model availability to India, South Korea, and Singapore","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/claude-region-expansion-in-sk/","posted_at":"2026-09-29T15:41:00Z","summary":"Anthropic's Claude Opus 5, Sonnet 5, and Haiku 4.5 expanded to Amazon Bedrock in India; Claude Opus 5 and Sonnet 5 to South Korea; and Claude Sonnet 5 to Singapore, with in-region or geographic cross-region inference for data residency.","theme":"bedrock-platform","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#38cd4deb-5fcb-427f-889a-64dc4de3f39d","title":"Amazon Rekognition Face Liveness now returns Feedback Codes","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/rekognition-liveness-feedback-codes/","posted_at":"2026-09-28T20:41:00Z","summary":"Amazon Rekognition Face Liveness now returns Feedback Codes in API responses to explain why a liveness check scored poorly, helping end users correct issues like poor lighting or face obstruction and retry successfully.","theme":"applied-ai","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-rekognition"]},{"id":"whats-new-v2#launches-e101f154-7670-4378-9019-79d8b452fc2a","title":"Grok 4.7 is now available on Amazon Bedrock","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/amazon-bedrock-grok-4-7/","posted_at":"2026-09-28T18:29:00Z","summary":"SpaceXAI's Grok 4.7 is now available on Amazon Bedrock, bringing improved mixed-document handling, repo-scale coding with planning and error recovery, and enhanced browser-use agents for form fills and portal navigation.","theme":"models","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#d3b0bc7d-b21e-4418-ac81-8c0f7cb6cae0","title":"Claude Sonnet 5.5 now available on AWS","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/claude-sonnet-5-5-aws/","posted_at":"2026-09-28T15:00:00Z","summary":"Anthropic's Claude Sonnet 5.5 is now available on Amazon Bedrock and Claude Platform on AWS, delivering stronger coding and knowledge-work capabilities at lower cost per task and faster speed than Sonnet 5.","theme":"models","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#78f24be7-9239-42af-b83e-14858a47d491","title":"AWS End User Messaging and Amazon SES now offer AI agent skills for the AWS MCP Server","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/aws-messaging-ses-ai-skills-mcp-server/","posted_at":"2026-09-25T07:00:00Z","summary":"AWS End User Messaging and Amazon SES now publish MCP server skills that let AI coding agents like Claude Code, Codex, and Cursor verify sending identities, build branded RCS agents, and send emails via natural language commands.","theme":"developer-tools","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-ses"]},{"id":"whats-new-v2#f168a41f-c0af-43cd-9fbd-692ee2790b58","title":"Amazon SageMaker HyperPod Inference Gateway for scalable LLM inference","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/sagemaker-hyperpod-inference-gateway/","posted_at":"2026-09-24T08:00:00Z","summary":"Amazon SageMaker HyperPod Inference Gateway is now available as an EKS managed add-on, using real-time inference-signal-driven routing to reduce first-token latency by up to 82% and p99 TTFT by 97–98% without requiring application code changes.","theme":"sagemaker-ml","significance":3,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-sagemaker-hyperpod","amazon-eks"]},{"id":"whats-new-v2#b2a08522-9e8f-42de-a622-76789c63a7be","title":"Amazon Bedrock Managed Knowledge Base now supports Salesforce and Zendesk as native data source connectors","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/amazon-bedrock-managed-knowledge-base-salesforce-zendesk-native-data-source-connectors/","posted_at":"2026-09-23T17:41:00Z","summary":"Amazon Bedrock Managed Knowledge Base added native Salesforce and Zendesk connectors that automatically crawl, extract metadata, and incrementally sync support articles and community posts into RAG-powered agents without custom ingestion pipelines.","theme":"bedrock-platform","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock","amazon-quick-suite","aiml"]},{"id":"whats-new-v2#gpt6solluna","title":"OpenAI GPT-6 Sol and GPT-6 Luna are now generally available on Amazon Bedrock","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/openai-gpt-6-sol-luna-on-amazon-bedrock/","posted_at":"2026-09-22T18:09:00Z","summary":"OpenAI's GPT-6 Sol and GPT-6 Luna are now generally available on Amazon Bedrock, giving teams a daily-workhorse coding model and a high-volume efficient model respectively, both supporting up to 1M tokens of context.","theme":"models","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#6399a6ad-560c-4906-b690-0fac30cb0e2c","title":"Claude Opus 5.5 is now available on AWS","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/claude-opus-5-5-aws/","posted_at":"2026-09-22T15:00:00Z","summary":"Anthropic's Claude Opus 5.5 is now available on Amazon Bedrock and Claude Platform on AWS, offering improved long-running coding and knowledge-work capabilities with adaptive reasoning effort, lower cost per token, and zero data retention support on Bedrock.","theme":"models","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#8b59e9f2-748d-4f73-9519-fa851af07d60","title":"AWS Security Hub AI Inventory adds Azure self-hosted instance support","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/security-hub-ai-inventory-azure-support/","posted_at":"2026-09-22T08:00:00Z","summary":"AWS Security Hub AI Inventory now discovers and catalogs AI assets on self-hosted Microsoft Azure instances using SBOM analysis, extending multi-cloud visibility over inference endpoints and AI agents to Azure VMs.","theme":"other","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["aws-security-hub","aiml"]},{"id":"whats-new-v2#product-wtaq4jwqxgc9y2dv9lttn5","title":"Kimi K3 by Moonshot AI is now generally available on Amazon Bedrock","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/moonshot-ai-kimi-k3-on-amazon-bedrock/","posted_at":"2026-09-18T15:15:00Z","summary":"Moonshot AI's Kimi K3, a 2.8-trillion-parameter open-weight model with native vision and a 1M-token context window, is now generally available on Amazon Bedrock as the first open-weight model there to support explicit prompt caching.","theme":"models","significance":3,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#96ec5087-284b-43d5-84be-9d264fc58181","title":"The new AgentCore Runtime is now available in Amazon Bedrock AgentCore","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/new-agentcore-runtime-generally-available","posted_at":"2026-09-18T13:25:00Z","summary":"Amazon Bedrock AgentCore Runtime V2 is now available with elastic memory reclamation and snapshot-based cold starts that hold P75 startup times to roughly 2 seconds even for 2 GB container images, cutting previous V1 times from 5–30 seconds.","theme":"agents","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-bedrock"]},{"id":"whats-new-v2#ea05bd09-8ee9-4347-a5b5-50685001bee1","title":"AWS HealthOmics now supports IAM session policies","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/omics-iam-session-policy/","posted_at":"2026-09-17T18:32:00Z","summary":"AWS HealthOmics now supports IAM session policies that let operators restrict permissions for individual workflow runs to specific S3 buckets or resources without creating dedicated IAM roles per tenant.","theme":"other","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-omics"]},{"id":"whats-new-v2#qwen3-6-35b-a3b-nvfp4-and-wan2-1-t2v-1-3b-diffusers-models-now-available-on-amazon-sagemaker-jumpstart","title":"Qwen3.6-35B-A3B-NVFP4 and Wan2.1-T2V-1.3B-Diffusers models now available on Amazon SageMaker JumpStart","url":"https://aws.amazon.com/about-aws/whats-new/2026/01/qwen3.6-35b-a3b-nvfp4-wan2.1-t2v-1.3B-diffusers-jumpstart/","posted_at":"2026-09-14T17:39:00Z","summary":"NVIDIA's Qwen3.6-35B-A3B-NVFP4 agentic coding model and Alibaba's Wan2.1-T2V-1.3B lightweight text-to-video diffusion model are now available on SageMaker JumpStart.","theme":"models","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["aiml","amazon-sagemaker-jumpstart"]},{"id":"whats-new-v2#ministral-3-3b-instruct-2512-and-ministral-3-8b-instruct-2512-models-now-available-on-amazon-sagemaker-jumpstart","title":"Ministral-3-3B-Instruct-2512 and Ministral-3-8B-Instruct-2512 models now available on Amazon SageMaker JumpStart","url":"https://aws.amazon.com/about-aws/whats-new/2026/01/ministral-3-3b-instruct-2512-ministral-3-8B-Instruct-2512-jumpstart/","posted_at":"2026-09-14T17:35:00Z","summary":"Mistral AI's Ministral-3-3B-Instruct-2512 and Ministral-3-8B-Instruct-2512 compact vision-capable models are now available on SageMaker JumpStart for edge and resource-constrained deployments.","theme":"models","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["aiml","amazon-sagemaker-jumpstart"]},{"id":"whats-new-v2#gemma-4-31b-it-assistant-and-gemma-4-31b-it-nvfp4-models-now-available-on-amazon-sagemaker-jumpstart","title":"Gemma-4-31B-it-assistant and Gemma-4-31B-IT-NVFP4 models now available on Amazon SageMaker JumpStart","url":"https://aws.amazon.com/about-aws/whats-new/2026/01/gemma-4-31b-it-assistant-gemma-4-31b-it-nvfp4-jumpstart/","posted_at":"2026-09-14T17:32:00Z","summary":"Google's Gemma-4-31B-it-assistant and NVIDIA's NVFP4-quantized variant are now available on SageMaker JumpStart, offering a 31B dense multimodal model with a 256K context window and function-calling support alongside a memory-efficient quantized option.","theme":"models","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-sagemaker-jumpstart","aiml"]},{"id":"whats-new-v2#granite-speech-4-1-2b-kanana-2-30b-a3b-instruct-and-openfold3-models-now-available-on-amazon-sagemaker-jumpstart","title":"granite-speech-4.1-2b, kanana-2-30b-a3b-instruct, and OpenFold3 models now available on Amazon SageMaker JumpStart","url":"https://aws.amazon.com/about-aws/whats-new/2026/01/granite-speech-4.1-2b-edge-kanana-2-30b-a3b-instruct-openfold3-jumpstart/","posted_at":"2026-09-14T17:30:00Z","summary":"IBM's granite-speech-4.1-2B multilingual ASR model, Kakao's kanana-2-30b-a3b Korean-English agentic model, and the OpenFold Consortium's OpenFold3 biomolecular structure prediction model are now available on SageMaker JumpStart.","theme":"models","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-sagemaker-jumpstart","aiml"]},{"id":"whats-new-v2#amazon-sagemaker-hyperpod-now-supports-model-caching-for-faster-inference-autoscaling-and-reduced-cold-starts","title":"Amazon SageMaker HyperPod now supports model caching for faster inference autoscaling and reduced cold starts","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/sgm-hyperpod-model-caching-inf/","posted_at":"2026-09-11T18:25:00Z","summary":"Amazon SageMaker HyperPod now supports model caching that pre-loads model weights onto local NVMe and pre-pulls container images, cutting inference cold starts by roughly 60% and eliminating over two minutes of image-pull time during autoscaling events.","theme":"sagemaker-ml","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-sagemaker"]},{"id":"whats-new-v2#3c649f78-f43f-41cc-8095-a83ecd5a59c0","title":"Amazon Bedrock Managed Knowledge Base now supports multimodal embeddings for video, audio, and image content with TwelveLabs Marengo 3.0","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/amazon-bedrock-managed-knowledge-base-multimodal-embeddings-twelvelabs-marengo/","posted_at":"2026-09-11T15:13:00Z","summary":"Amazon Bedrock Managed Knowledge Base now integrates TwelveLabs Marengo 3.0 to generate 512-dimensional multimodal embeddings directly from video, audio, and image content, enabling semantic search across media without transcription and unlocking use cases like sports analytics and contextual lecture retrieval.","theme":"bedrock-platform","significance":2,"fetched_at":"2026-10-28T01:58:04.998Z","services":["amazon-quick-suite","amazon-quicksight","aiml"]},{"id":"whats-new-v2#26e35036-cc41-42dc-91ce-ba7c2f73024f","title":"AWS HealthOmics now publishes real-time run metrics to Amazon CloudWatch","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/aws-healthomics-realtime-run-metrics/","posted_at":"2026-09-11T15:00:00Z","summary":"AWS HealthOmics added 14 real-time run metrics to Amazon CloudWatch, letting bioinformatics teams observe CPU, GPU, memory, and I/O utilization during workflow execution to right-size resources and preempt failures.","theme":"other","significance":1,"fetched_at":"2026-10-28T01:58:04.998Z","services":[]}]}