AWS News - 2026-11-10

Amazon DocumentDB Serverless now provides up to 35% higher throughput with the launch of platform version 4. Platform versions represent improvements in performance, scaling capabilities, or features, and are automatically managed at the cluster level.

All new serverless clusters launch on the latest platform version available in your AWS Region. You can upgrade existing clusters on platform version 3 to platform version 4 by applying the serverless-platform-version-update pending maintenance action. This gives you control over when to apply the update, either immediately or during your next scheduled maintenance window. After you upgrade to a newer platform version, you can't downgrade to a previous version.

You can view the platform version of a cluster in the AWS Management Console or through the ServerlessV2PlatformVersion field of the DescribeDBClusters API. For more information, see Amazon DocumentDB serverless scaling capacity and Checking and upgrading the platform version of a serverless cluster.

Today, AWS is introducing an open-source solution with a local web UI that helps customers turn data lake records into graphs for analysis with Amazon Neptune Analytics. By projecting selected records and relationships, customers can investigate connections across tables and run graph algorithms without building their own ETL pipelines.

For example, a customer could project accounts and transactions into a graph to explore how accounts are connected or identify communities that may warrant investigation. Doing this today often means writing queries to prepare node and relationship data, creating a Neptune Analytics graph, and managing the import. Customers may also need to build a selective projection so they do not load an entire dataset for a narrower question.

The new UI brings those steps into single workflow. To run graph analytics over iceberg tables, customers select an Athena catalog and database, define a property graph using SQL/PGQ, and preview rows before import. The solution runs the projection queries through Athena, stages the selected data in Amazon S3, and imports it into Neptune Analytics. The UI helps customers manage the resulting graph. They can open the loaded graph in Graph Explorer to query and visualize relationships. [1]

The UI runs locally and it is available in all regions where Amazon Neptune Analytics is available. To get started, see the setup instructions.

Amazon Connect Customer now provides evaluation form templates so managers don't have to build forms from scratch, reducing the time to deploy automated evaluations of human and AI agent performance. These templates cover typical customer interaction scenarios: customer service, inbound sales, and self-service. Each template assesses human and AI agent behaviors. For example, the customer service template includes questions such as "Did the agent show empathy toward the customer's issue?" Templates also include questions that assess business outcomes such as customer satisfaction and attrition risk, so managers can see how agent behaviors relate to outcomes. Managers can use the templates as-is or customize them before deploying.

This feature is available in the following AWS Regions: US East (N. Virginia), US West (Oregon), Canada (Central), Europe (Frankfurt), Europe (London), Asia Pacific (Singapore), Asia Pacific (Sydney), and Asia Pacific (Tokyo). To learn more, visit our documentation and our webpage. For Amazon Connect Customer pricing, visit our pricing page.

Amazon Managed Workflows for Apache Airflow (Amazon MWAA) now supports DuckDB operators. DuckDB is an embedded analytical database that lets customers perform data quality checks, summaries, and transformations on data from sources DuckDB supports, including Amazon S3.

Native support is available for DuckDB on Amazon MWAA Provisioned environments running Apache Airflow 3.3.1. DuckDB comes preinstalled and preconfigured for AWS connections in the latest MWAA Airflow 3.3.1 update. After updating their MWAA environment to Airflow 3.3.1 and adding execution role access to their data stores, customers can start writing tasks with DuckDBExecuteQueryOperator with no additional setup.

This feature is available in all AWS Regions where Amazon MWAA is available. For more information, see the AWS Region table.

To learn more, see Using DuckDB with Amazon MWAA in the Amazon MWAA User Guide.

Amazon Bedrock, the platform for building generative AI applications and agents at production scale, now offers Claude Mythos 5.2 to eligible customers in AWS GovCloud (US). Claude Mythos 5.2 is Anthropic's most advanced AI model to date for use cases across cybersecurity, software coding, and complex reasoning tasks. The model can identify sophisticated security vulnerabilities in software and demonstrate exploitability. It can also comprehend large codebases and deliver actionable findings with less manual guidance than previous AI models. This enables security teams to accelerate defensive cybersecurity work, find and fix security vulnerabilities in the world's most critical software, and address them before threats emerge. Claude Mythos 5.2 is the first of Anthropic's AI models with powerful cybersecurity capabilities to be available in AWS GovCloud (US).

Claude Mythos 5.2 is available to eligible customers in AWS GovCloud (US) Regions through Amazon Bedrock. Access is limited to allow-listed organizations; to get started, visit Anthropic's Cyber Verification Program. 

Amazon Quick, the AI workspace built only for work, is now AWS Quick, a name that reflects the enterprise security and governance built into Quick from the start as an AWS service. The new name now appears across the Quick website, the web, desktop, and mobile apps, and AWS documentation. Existing customers don't need to take any action, and Quick works the same way it did before.

Amazon Quick Sight, the business intelligence capability in Quick, is now AWS Quick Sight. Quick Sight customers can keep using their dashboards, datasets, and analyses just as they do today, with no changes to their data, settings, or pricing.

With Quick, teams use AI agents to research, analyze, and act across the data and applications they already use, regardless of vendor. Quick understands your role, your team, and how your work connects, and anyone can build agents, apps, and automations in plain language that become shared assets for the whole team. Quick is HIPAA eligible and holds compliance certifications including FedRAMP, SOC 2, and ISO 27001. Your data stays in your environment and never trains someone else's models.

AWS Quick is available in many AWS Regions. For more information, see the AWS Region table.

To learn more, see AWS Quick.

AWS Glue now supports 100 new native connectors that simplify access to data in popular SaaS and REST API-based applications. Previously, bringing data from SaaS applications into AWS required writing custom code. Now, teams can create a connection once in the AWS Glue Data Catalog and reuse it across ETL jobs with the same permissions model they already use for AWS Glue.

The new connectors span generative AI, finance, payments, marketing, product analytics, human resources, engineering, IT operations, and collaboration tools. AI teams can bring usage and cost data from model providers such as Anthropic to track generative AI adoption and spend. Finance teams can consolidate transactions from applications such as Adyen, Square, and Xero for faster reporting. Product and marketing teams can combine billing, usage, and campaign data from tools such as Chargebee, Amplitude, and Klaviyo to understand customer lifetime value. HR teams can unify workforce data from systems such as BambooHR and Gusto for headcount planning. Engineering teams can join incident and deployment data from tools such as PagerDuty, Sentry, Jenkins, and Vercel to track reliability. You can load the integrated data into Amazon S3 or Amazon Redshift for analytics and reporting.

The 100 new AWS Glue connectors are available in all AWS Commercial Regions on Glue versions 5.0 and above.

To learn more, see the complete list of available connectors and learn about connecting to individual connectors in the AWS Glue documentation.

Amazon CloudWatch Logs now supports account-level retention policies, so you can use a single policy to set how long log data is kept across all or subset of log groups in an AWS account. Account-level policies allow platform and security teams to bring each log group in line with their retention standards.

With today's launch, you can create policies that cover all log groups in the account or apply to a specific set of log groups with names that start with a prefix, such as /aws/lambda. You can additionally choose whether it applies to existing log groups or only new ones. When more than one policy matches a log group, the longest retention period applies. For example, a central platform team can set a 90-day baseline for the account and a 1-year policy for log groups that start with /security to meet audit requirements, without tracking individual log group names.

Account-level retention policies are available in all commercial AWS Regions except Middle East (Bahrain), and Middle East (UAE) at no additional cost . You can create and manage up to 20 policies per account and region using the AWS Management Console, AWS CLI, and AWS SDKs. Learn more in the  CloudWatch Logs documentation, and the  API reference .

Today, AWS announces the general availability of the Web Search API on Amazon Bedrock AgentCore, which lets you search the web with a single API call. Web Search on AgentCore has been available as an MCP tool through AgentCore Gateway. The new InvokeWebSearch API gives applications, custom agent loops, and data pipelines search results with no resources to create.

The API returns ranked results from the Amazon-operated web index behind Web Search on AgentCore. Each result includes a text extract and, when available, the page title, URL, and publication time. You can filter each request by domain and publication time, for example to find recent product announcements or check a claim against trusted domains.

You call the API with the AWS SDKs or the AWS Command Line Interface (AWS CLI) and authorize calls with AWS Identity and Access Management (IAM). You can call it from your virtual private cloud (VPC) through AWS PrivateLink and log requests as AWS CloudTrail data events. Web Search API queries are billed at $7 per 1,000 queries, the same rate as Web Search through AgentCore Gateway, with no gateway charges.

The Web Search API is available in US East (N. Virginia), US East (Ohio), US West (Oregon), Asia Pacific (Sydney), Asia Pacific (Tokyo), and Europe (Ireland). To learn more, see Search the web with the InvokeWebSearch API in the AgentCore Developer Guide, the Web Search  page, and AgentCore pricing .

Amazon DocumentDB (with MongoDB compatibility) 8.0.3 adds hybrid post-quantum key exchange for TLS, wildcard indexes, the $unionWith aggregation stage, and new change stream events. Post-quantum TLS protects data in transit from traffic that is recorded today and decrypted later with a quantum computer. The new MongoDB APIs let you move more MongoDB workloads to Amazon DocumentDB without changing application code.

What's new in 8.0.3:

• Post-quantum TLS: TLS 1.3 connections use hybrid key exchange based on the Module-Lattice-Based Key-Encapsulation Mechanism (ML-KEM) when the client supports it. Other clients keep using classical key exchange. Two new Amazon CloudWatch metrics, TLSHandshakesClassical and TLSHandshakesPostQuantum, count the handshakes of each type.

• Wildcard indexes: index every field, or every field under a path, to query documents with dynamic schemas.

• Aggregation: $unionWith combines results from multiple collections in one pipeline. $lookup and $graphLookup work in views, and correlated $lookup sub-pipelines accept $unionWith and $facet.

• Change streams: dropIndex() and dropIndexes() now emit dropIndexes events.

• Compression metrics: collStats and stats() report per-collection input bytes, output bytes, and compression ratio.

Amazon DocumentDB 8.0.3 is available in all AWS Regions where Amazon DocumentDB is offered. To learn more, see the Amazon DocumentDB Developer Guide.

Amazon OpenSearch Service now supports clone-based upgrades, which let you validate a new OpenSearch version alongside your current one and cut over to it only when you're satisfied. If you find a problem, you can cancel before you cut over or roll back afterward. With clone-based upgrades, you can move to new versions with more confidence.

When you start a clone-based upgrade, OpenSearch Service copies the data from your existing cluster (the primary cluster) to a second cluster running the new version (the cloned cluster). While the primary cluster keeps serving your workload, you can run the same search queries on both clusters and compare their performance using Amazon CloudWatch metrics. After you finish validating, you can cut over to the cloned cluster. If you keep the previous version for a retention period you choose, you can roll back to it at any time during that period. You can start and manage a clone-based upgrade from the AWS Management Console, the AWS CLI, or the OpenSearch Service APIs.

Clone-based upgrades are available for domains running OpenSearch 1.1 or later, or Elasticsearch 7.10, in AWS Commercial Regions where Amazon OpenSearch Service is offered, except Asia Pacific (New Zealand), Asia Pacific (Taipei), Asia Pacific (Thailand), and Mexico (Central). For a complete list of Regions, see the AWS Region Table.

To learn more, see Upgrading Amazon OpenSearch Service domains in the Amazon OpenSearch Service Developer Guide. To get started, visit the Amazon OpenSearch Service product page.

Amazon CloudWatch alarms that use Prometheus Query Language (PromQL) can now trigger Amazon EC2 and Application Auto Scaling step and simple scaling policies. You can scale your EC2 Auto Scaling groups on CloudWatch OpenTelemetry metrics using the same kind of PromQL queries you use for dashboards and alerting.

In this release, PromQL alarms also support EC2 Auto Scaling simple and step scaling policies. Target tracking is not currently a supported policy type for PromQL alarms. Teams moving to OpenTelemetry can scale capacity on the same metrics as the rest of their monitoring, instead of keeping separate classic metric alarms for scaling. For step scaling, how far the value is past the threshold in your PromQL query decides which step applies. The query must return a single series and compare it to a threshold.

You can add EC2 Auto Scaling and Application Auto Scaling actions in the CloudWatch console when you create or edit an alarm, or with the PutMetricAlarm API, AWS CLI, AWS SDKs, or AWS CloudFormation. You can also use the PutMetricAlarm API, AWS CLI, AWS SDKs, or AWS CloudFormation.

This feature is available in all AWS Regions where CloudWatch PromQL alarms are available. Standard CloudWatch alarm pricing applies. To get started, see  Using PromQL in alarms  and  Alarm actions  in the Amazon CloudWatch User Guide.

Amazon S3 Tables now support server access logging, providing you detailed records of the requests made to your table buckets. You can now track table operations, namespace management, and data access to support security audits, access analysis, and cost attribution. Log delivery is available at no additional cost, and you pay only for the storage of the log files.

When you enable server access logging on a table bucket, S3 delivers logs for requests to all tables and namespaces in that bucket to a general purpose bucket in the same account and Region. Each record captures the requester, operation, table name, namespace, response status, and request timing, using the same space-delimited format as S3 server access logs for general purpose buckets. This lets you reuse your existing log analysis workflows to query S3 Table logs with any query engine of your choice such as Amazon Athena and DuckDB.

Server access logging for S3 Tables is available in all AWS Regions where S3 Tables are available. You can enable it using the AWS Management Console, AWS CLI, AWS SDKs, or AWS CloudFormation. To learn more, visit the S3 Tables overview page and S3 Tables server access logging in the Amazon S3 User Guide.

Today, AWS Marketplace announces that free trials of usage-based software as a service (SaaS) products can now convert automatically to paid subscriptions. When a free trial ends, the buyer's subscription continues at the seller's usage-based pricing, so buyers pay only for what they use after the trial. This helps sellers turn more trial users into paying customers. Buyers get a simple way to evaluate software before they buy, with no new subscription and no interruption in access.

Sellers can offer a free trial of 7 to 90 days that converts to a paid subscription at the price they set. For example, a data team can try an analytics product for 30 days and, if it meets their needs, keep using it after the trial ends. Buyers see the trial length, conversion date, and pricing before they subscribe. They receive reminder emails before the trial converts and can cancel at any time before conversion at no charge.

Sellers can create a free trial in AWS Partner Central, or update an existing free trial offer to convert automatically. This capability is available for public offers in all AWS Regions where AWS Marketplace is available.

To learn more, see the AWS Marketplace Seller Guide and the AWS Marketplace Buyer Guide.

Amazon GameLift Servers now offers Game Session Placement Insights, delivering detailed, per-placement diagnostic logs for game session queue routing. Amazon GameLift Servers routes game session placement requests through queues that evaluate multiple fleet locations based on latency and cost factors. This feature is designed for game developers and engineers who need greater transparency into how their game session queues operate.

Previously, developers had limited visibility into why placements succeeded, failed, or timed out, making queue optimization difficult. Game Session Placement Insights captures which fleet locations were evaluated, the player latency and cost factors that influenced routing decisions, and the final placement outcome for every placement. This enables teams to identify patterns such as consistently slow placements, repeated failures, or sessions landing in unexpected locations, and to tune their queues to improve the overall player experience.

You can enable this feature per queue and deliver logs to Amazon CloudWatch Logs, Amazon S3, and Amazon Data Firehose. Game Session Placement Insights is available in all supported Amazon GameLift Servers regions.

To learn more about Game Session Placement Insights and how to get started, visit the Amazon GameLift Servers documentation.

Eligible customers can now access OpenAI’s specialized cybersecurity models Daybreak Red: GPT-5.6 Cyber and Daybreak Blue: GPT-5.6 Sol through the bedrock-runtime endpoint on Amazon Bedrock.

Daybreak Blue is the starting point for most security teams across defensive workflows, including vulnerability discovery, detection engineering, and incident response. Daybreak Red is designed for advanced, authorized tasks such as vulnerability research, exploit reproduction, and mitigation development. For these tasks, a lower refusal threshold, matched by stronger identity verification, monitoring, and access controls, improves the speed and depth of an investigation. Your inference data is not used for model training, and neither model requires you to opt in to sharing your data with OpenAI.

On the bedrock-runtime endpoint, the models are available through US and Global cross-Region inference (CRIS) profiles. Access to the models requires enrollment in Daybreak access from OpenAI. To enroll, contact OpenAI or reach out to your AWS account team for guidance on eligibility. Once approved, work with your account team to request access on AWS. For regional availability through the bedrock-mantle endpoint and more information, view the documentation.

Amazon S3 Access Grants can now discover and list all your grants across multiple AWS accounts with a single API call, eliminating the need to query each account separately. A single ListCallerAccessGrants call returns grants from your own S3 Access Grants instance and every instance shared with you, giving you a complete view of the S3 data you can access.

By setting the new InstanceScope parameter to ALL in your ListCallerAccessGrants request, S3 Access Grants queries every instance shared with you through AWS Resource Access Manager (AWS RAM) and returns all the grants in one combined response, each identifying the account that owns it. Additionally, when you request credentials with GetDataAccess for an S3 path and set InstanceScope to ALL, S3 Access Grants locates the owning account for you, routes the request to its instance, and returns credentials for the matching grant.

This capability is now available in all AWS Regions where S3 Access Grants is available, at no additional cost. You can use the AWS SDK, API, or CLI to list and access grants across accounts. To learn more, visit the S3 Access Grants overview page and the S3 documentation.

Most of the cost of adopting something new isn’t the new thing, it’s the move. Moving a Hive table to Apache Iceberg has typically meant weeks of engineering work. Last week, Amazon S3 Tables turned that into three steps: start, sync as often as you need, complete. The source stays writable except for a few […]

Amazon Athena にシステムテーブルが加わり、第一弾として query_executions の提供が始まりました。すべてのワークグループで実行したクエリの記録が Amazon S3 Tables 上の Apache Iceberg テーブルに自動で保存され、Athena から直接クエリできます。本記事では、システムテーブルの概要と始め方に加え、コストの配分、パフォーマンス分析、監査、運用監視、キャパシティ予約の最適化に使えるクエリの例を紹介します。

30 以上の言語に対応する製品サポートチャットを Amazon Bedrock AgentCore で実現した、国内大手製造業の事例です。Runtime V2 による起動の高速化、会話文脈の保持、回答品質の評価の工夫を紹介します。

AI エージェントの中核にある大規模言語モデル (LLM) は非決定論的で、ハルシネーションやプロンプトインジェクションのリスクがあります。本記事では、LLM を信頼できないアクターとして扱う考え方を説明します。そのうえで、Amazon Bedrock AgentCore Gateway でツール呼び出しを認可制御する AgentCore Policy が Cedar を採用した理由を解説します。可読性、決定論的な動作、自動推論による分析可能性を活かして、ニューロシンボリック AI でポリシーを作成し、Cedar Analysis で論理エラーや競合を検出する仕組みも紹介します。

Amazon Bedrock を活用した生成 AI アプリケーションでは、設計段階から安全性を組み込むことが重要です。本記事では、責任ある AI の主要な側面と、設計と開発、デプロイ、運用の各フェーズで安全性を確保するための考え方を解説します。さらに、Amazon CloudWatch による監視や Amazon Bedrock Guardrails による保護措置のカスタマイズなど、不正使用を防止する方法を紹介します。AWS Trust & Safety チームから不正使用の報告を受けた際の対応プロセスも説明します。

AWS re:Invent 2026 (11 月 30 日〜12 月 4 日、ラスベガス) では、AWS Transform のエージェント型 AI が移行ライフサイクル全体をどのように自動化するかを紹介するセッションが多数用意されています。本記事では、移行戦略とお客様事例、ビジネスケースの構築、AWS Transform と Amazon EVS のハンズオンセッション、AWS Outposts への移行、そして移行後のモダナイゼーションまで、インフラ移行に関するセッションを厳選して紹介します。

2026 年 10 月 26 日に開催したセキュリティセミナーの開催報告です。「防御・検知」「説明責任」「復旧」の 3 つの軸で、AWS とパートナー 5 社のセッション内容と、軸ごとにまず着手したい対策をまとめました。

AWS re:Invent 2026 では、AWS 上で Microsoft ワークロードを実行しているお客様向けに充実したセッションを用意しています。AWS Transform と Kiro によるエージェンティック AI での .NET と SQL Server のモダナイゼーション、ライセンスコストの削減、VMware からの移行、レジリエンスの高い運用まで、ぜひ参加していただきたいセッションを紹介します。

Amazon Web Services (AWS) を活用している組織は、セキュリティとコスト最適化に関する推奨事項を毎月数千件も受け取ります。しかし、優先順位の問題やリソースの制約からその多くは実施されないままとなっています。AWS Managed Services (AMS) の Trusted Remediator はこの課題を解決するため、複数の AWS アカウントにわたる修復を自動化し、手作業による修復にかかる時間と手間を大幅に減らします。あらかじめ用意された修復ライブラリは、116 件の推奨事項に対応しており、新しい機能が加わるたびに増え続けています。これらの自動化は、セキュリティ、コスト最適化、運用上の優秀性の各分野にわたります。企業はセキュリティホールを効率よくふさぎ、コストを最適化し、コンプライアンスを維持しながら AWS のベストプラクティスの変化にも柔軟に対応できます。

みなさん、こんにちは。AWS ジャパン 執行役員 パブリックセクター技術統括本部長の瀧澤です。 電子カルテベン […]

Amazon Elastic Kubernetes Service (Amazon EKS) 向けに AWS Nitro Sandboxes の提供を開始しました。サンドボックス化されたポッドは AWS Nitro System を基盤とする専用の軽量 VM で実行されるため、ポッドごとにハードウェアベースのセキュリティ境界を設けられます。EKS Auto Mode ではポッド仕様を 1 行変更するだけで有効にでき、分離モードと機密モードを選べます。AI エージェント、マルチテナント SaaS、規制対象のワークロードを安全かつ効率的に実行する方法を、AI コーディングエージェントの実行例とともに紹介します。

You can now use tokens from AWS Identity and Access Management (IAM) outbound federation to assume roles in other Amazon Web Services (AWS) partitions. When your workloads run across AWS partitions such as Commercial, China, GovCloud (US), and European Sovereign Cloud, a workload in one partition might need to access resources or communicate with AWS […]

Amazon Bedrock AgentCore Memory helps AI agents maintain context across sessions through persistent short-term and long-term memory. Because memory content is automatically retrieved and injected into future conversations, the integrity of what gets written to memory directly shapes future agent behavior, which makes the write path a valuable control point to strengthen. AgentCore already provides […]

Bulletin ID: 2026-174-AWS
Publication Date: 11/09/2026 13:00 PM PST

Description:

AWS ParallelCluster CLI is an open source command line tool, distributed on PyPI as aws-parallelcluster. We identified CVE-2026-124033, an issue where the export-cluster-logs and export-image-logs commands in the AWS ParallelCluster CLI before version 3.16.0 contain a relative path traversal issue. An authenticated remote user with write access to the log export Amazon S3 bucket could use a crafted S3 object key to write files outside the intended export directory on the machine running the command.

Impacted versions: >= 3.0.0 AND < 3.16.0

Refer to the article for the most up-to-date and complete information.

Read this article

Bulletin ID: 2026-175-AWS
Publication Date: 11/09/2026 14:15 PM PST

Description:

Amazon Braket SDK is an open source Python library that customers use to run quantum tasks and hybrid jobs on Amazon Braket.

We identified CVE-2026-124034, an issue in the local hybrid job input data download in the Amazon Braket SDK. And we identified CVE-2026-124035, an issue in the error mitigation schema validation in Amazon Braket Schemas, the library the Amazon Braket SDK uses to parse quantum task results.

Impacted versions:
- amazon-braket-sdk >= 1.10.0 AND < 1.127.7
- amazon-braket-schemas >= 1.17.0 AND < 1.32.2

Refer to the article for the most up-to-date and complete information.

Read this article

DeepL runs a multi-Region hybrid network across multiple AWS Regions and AWS Direct Connect locations. See how DeepL drains a single Direct Connect link for planned maintenance without packet loss or cross-Region rerouting, using AWS Cloud WAN and an on-premises AS_PATH prepend so every other Region keeps its closest-exit routing.

Learn how Novus Hi-Tech, with the AWS Prototyping and AI Customer Engineering (PACE) team, built FleetGPT, an agentic AI solution on Amazon Bedrock that turns fleet IoT data into real-time safety interventions through natural language querying, automated investigation, and AI-powered driver voice calls, cutting road-behavior violations by 46% and drowsiness alerts by 55%.

Introducing AWS Quick, formerly Amazon Quick, and the debut of our first brand campaign celebrating teams who keep business moving. AWS Quick is the AI workspace built only for work, where teams use AI agents to research, analyze, and act across the data and applications they already rely on.

CrossnoKaye brought physical AI to heavy industry with Amazon Bedrock AgentCore. Its ATLAS platform and AURA agentic system ground AI reasoning in live plant data and run verifiable Python in isolated AgentCore sandboxes, giving plant engineers checkable, math-backed answers instead of model guesses.

A seven-layer composable reference architecture for taking enterprise agentic AI from prototype to production on Amazon Bedrock AgentCore. Each layer has a stable contract, so you can swap technologies at any layer without redesigning the rest of the stack.

NVIDIA Nemotron 3.5 Super VL, a 120B-parameter multimodal vision-language model, is now available in Amazon SageMaker JumpStart. Deploy the NVFP4 or BF16 variant in a few steps, then run text, image, and video inference for document intelligence, computer-use agents, and multimodal agentic workloads.

Claude Fable 5.2 and Claude Mythos 5.2 are now available on Amazon Bedrock and Claude Platform on AWS. Learn what's new in both models, why Fable 5.2 matters for your technical workloads, and how to get started building with it on Amazon Bedrock.

Amazon Bedrock Managed Knowledge Base now supports single-shot RAG and retrieval-only requests through AgenticRetrieveStream, alongside agentic retrieval. This post explains how the maxAgentIteration parameter matches retrieval effort to each request, using an illustrative internal banking policy assistant.

Repeatable tasks like status updates, release notes, and ticket creation can be encoded in AWS Quick through agents and skills. This post covers how agents, skills, and your data fit together in AWS Quick, how to build and share skills for your team, and how the official catalog and custom agents help you use them.