AWS Client VPN introduces a rebuilt AWS VPN Client v6.0.x which offers new features like command-line interface (CLI) support, enterprise administrative controls, and faster connection establishment time, making it easier you to automate VPN connectivity and centralize device management across your organization.

The AWS VPN Client CLI provides full feature parity with the GUI. You can now script VPN connections into your automation workflows and infrastructure-as-code deployments. Previously, integrating VPN connectivity into automated environments required third-party tooling or manual intervention. This feature eliminates that by supporting background CLI operations. Previously, you had to distribute VPN profiles among all users in your organization, which could be managed by any user without permissions. Now, with administration controls on AWS client, you can centralize VPN policy enforcement by scoping profiles to specific users, manage global profiles available to all users on a device, and enforce approved VPN configurations across your organization.

The client is rebuilt with OpenVPN3, delivering faster connection establishment across all supported operating systems. You can use both the GUI and CLI together as both run concurrently and VPN connections persist independently of either interface. The rebuilt client v6.0 onwards maintains full backward compatibility with existing AWS Client VPN endpoints, so no endpoint changes are required. The updated AWS VPN Client is available today for Windows (x64/ARM), macOS (x64/ARM), and Linux (x64). There are no additional charges beyond standard pricing of AWS Client VPN. Download the latest client version 6.0.x for macOS, Windows and Linux to start using it. 

To learn more about Client VPN, visit the AWS Client VPN product page or read the documentation.

​AWS Client VPN introduces a rebuilt AWS VPN Client v6.0.x which offers new features like command-line interface (CLI) support, enterprise administrative controls, and faster connection establishment time, making it easier you to automate VPN connectivity and centralize device management across your organization. The AWS VPN Client CLI provides full feature parity with the GUI. You can now script VPN connections into your automation workflows and infrastructure-as-code deployments. Previously, integrating VPN connectivity into automated environments required third-party tooling or manual intervention. This feature eliminates that by supporting background CLI operations. Previously, you had to distribute VPN profiles among all users in your organization, which could be managed by any user without permissions. Now, with administration controls on AWS client, you can centralize VPN policy enforcement by scoping profiles to specific users, manage global profiles available to all users on a device, and enforce approved VPN configurations across your organization. The client is rebuilt with OpenVPN3, delivering faster connection establishment across all supported operating systems. You can use both the GUI and CLI together as both run concurrently and VPN connections persist independently of either interface. The rebuilt client v6.0 onwards maintains full backward compatibility with existing AWS Client VPN endpoints, so no endpoint changes are required. The updated AWS VPN Client is available today for Windows (x64/ARM), macOS (x64/ARM), and Linux (x64). There are no additional charges beyond standard pricing of AWS Client VPN. Download the latest client version 6.0.x for macOS, Windows and Linux to start using it.  To learn more about Client VPN, visit the AWS Client VPN product page or read the documentation.  

AWS GovCloud (US) now offers Claude Opus 5 — the most advanced Opus model yet, and compatible with zero data retention (ZDR) — bringing a step-change in coding, long-running agents, and complex professional work to teams building at the highest level. Claude Opus 5 is available via the bedrock-runtime endpoint in both AWS GovCloud (US) regions, and available via the bedrock-mantle endpoint in AWS GovCloud (US-West)

Claude Opus 5 delivers advances in coding, understanding and navigating codebases like an experienced engineer and writing production-quality code while adapting its strategy as it works. It powers dependable agents that run for hours and even overnight, finding paths around obstacles, recovering from errors, and reaching their objectives. And it brings deeper reasoning to long documents and higher accuracy to complex analysis, with the largest gains on document-heavy enterprise work. 

Amazon Bedrock offers Claude Opus 5 with zero data retention (ZDR) enabled by default, giving you Opus’ top-tier intelligence while meeting your data governance requirements. It keeps your data within AWS infrastructure with regional data residency and provides access through a unified service with AWS-managed features like Guardrails and Knowledge Bases. To learn more, see the Amazon Bedrock documentation and regional availability.

​AWS GovCloud (US) now offers Claude Opus 5 — the most advanced Opus model yet, and compatible with zero data retention (ZDR) — bringing a step-change in coding, long-running agents, and complex professional work to teams building at the highest level. Claude Opus 5 is available via the bedrock-runtime endpoint in both AWS GovCloud (US) regions, and available via the bedrock-mantle endpoint in AWS GovCloud (US-West)
Claude Opus 5 delivers advances in coding, understanding and navigating codebases like an experienced engineer and writing production-quality code while adapting its strategy as it works. It powers dependable agents that run for hours and even overnight, finding paths around obstacles, recovering from errors, and reaching their objectives. And it brings deeper reasoning to long documents and higher accuracy to complex analysis, with the largest gains on document-heavy enterprise work. 
Amazon Bedrock offers Claude Opus 5 with zero data retention (ZDR) enabled by default, giving you Opus’ top-tier intelligence while meeting your data governance requirements. It keeps your data within AWS infrastructure with regional data residency and provides access through a unified service with AWS-managed features like Guardrails and Knowledge Bases. To learn more, see the Amazon Bedrock documentation and regional availability.  

Today, Amazon Quick announces the general availability of Microsoft 365 extensions for Excel, PowerPoint, Word, and Outlook. These extensions enable Quick to perform tasks directly within users’ M365 environments, using AI to handle complex local tasks such as redlining documents, building financial models, creating presentation-ready decks, and managing Outlook inboxes.

The Excel extension helps with complex spreadsheet analysis, creating pivot tables and charts, and importing and cleaning data. The PowerPoint extension helps you create and refine presentations from Quick data using organization-defined templates. The Word extension generates formatted documents with Word primitives, makes sweeping edits with track changes enabled, and participates as a reviewer in comments. The Outlook extension performs inbox and calendaring tasks such as prioritizing emails, organizing your inbox, scheduling meetings, and drafting replies using your Quick data and entire inbox context.

These extensions transform daily work across teams. Finance teams can build complex models by describing what they need. Sales teams can draft proposals that automatically pull from CRM data. Marketing teams can create branded presentations without manual formatting. Legal teams can streamline contract reviews. Operations teams can manage email workflows and schedule meetings intelligently, and IT teams can automate routine data analysis that previously required manual effort.

Amazon Quick Microsoft 365 extensions are available in US East (N. Virginia), US West (Oregon), Asia Pacific (Sydney), Europe (Ireland), Asia Pacific (Tokyo), and Europe (Frankfurt). To learn more, see Amazon Quick for Microsoft 365: Agentic AI where you work, and download extensions on the Quick download page.

​Today, Amazon Quick announces the general availability of Microsoft 365 extensions for Excel, PowerPoint, Word, and Outlook. These extensions enable Quick to perform tasks directly within users’ M365 environments, using AI to handle complex local tasks such as redlining documents, building financial models, creating presentation-ready decks, and managing Outlook inboxes.
The Excel extension helps with complex spreadsheet analysis, creating pivot tables and charts, and importing and cleaning data. The PowerPoint extension helps you create and refine presentations from Quick data using organization-defined templates. The Word extension generates formatted documents with Word primitives, makes sweeping edits with track changes enabled, and participates as a reviewer in comments. The Outlook extension performs inbox and calendaring tasks such as prioritizing emails, organizing your inbox, scheduling meetings, and drafting replies using your Quick data and entire inbox context.
These extensions transform daily work across teams. Finance teams can build complex models by describing what they need. Sales teams can draft proposals that automatically pull from CRM data. Marketing teams can create branded presentations without manual formatting. Legal teams can streamline contract reviews. Operations teams can manage email workflows and schedule meetings intelligently, and IT teams can automate routine data analysis that previously required manual effort.
Amazon Quick Microsoft 365 extensions are available in US East (N. Virginia), US West (Oregon), Asia Pacific (Sydney), Europe (Ireland), Asia Pacific (Tokyo), and Europe (Frankfurt). To learn more, see Amazon Quick for Microsoft 365: Agentic AI where you work, and download extensions on the Quick download page.  

Today, AWS announces support for AWS Local Zones in Spot placement score, helping you identify locations where your Spot capacity request is most likely to succeed. Spot placement score evaluates your target capacity and compute requirements and returns scores for AWS Regions or Availability Zones.

Previously, Spot placement scores excluded local zone capacity information when reporting zonal and regional scores. Now, you can optionally request local zones to be included in the zonal and regional score responses giving you a broader view of your Spot capacity options.

You can use Spot placement score through EC2 Spot Console, AWS CLI, or SDK. To learn more about Spot placement score see Spot placement score documentation.

​Today, AWS announces support for AWS Local Zones in Spot placement score, helping you identify locations where your Spot capacity request is most likely to succeed. Spot placement score evaluates your target capacity and compute requirements and returns scores for AWS Regions or Availability Zones.
Previously, Spot placement scores excluded local zone capacity information when reporting zonal and regional scores. Now, you can optionally request local zones to be included in the zonal and regional score responses giving you a broader view of your Spot capacity options.
You can use Spot placement score through EC2 Spot Console, AWS CLI, or SDK. To learn more about Spot placement score see Spot placement score documentation.  

• Categoría: Company News

agosto 13, 2026

Cómo desplegar Fireworks AI en Microsoft Foundry: un plan de arquitectura para startups

Publicado en Microsoft para Startups.

Muchas startups comienzan a construir productos de IA con un único modelo cerrado como una manera rápida de prototipar. Pero la selección de modelos para cargas de trabajo de aplicaciones es una decisión que se acumula a lo largo del ciclo de vida de un producto y determina los costes futuros y la diferenciación del producto.

Los modelos abiertos les permiten tomar esa decisión de manera más deliberada: ustedes eligen el modelo, lo ajustan a su caso de uso y moldean el comportamiento que distingue a su producto. Con Fireworks AI en Microsoft Foundry ya disponible a nivel general, pueden ofrecer inferencia de modelos abiertos de alto rendimiento y baja latencia directo en Azure. No necesitan construir su propia infraestructura de inferencia para ejecutar modelos abiertos aquí; Fireworks les sirve en Foundry, así que pueden empezar rápido y escalar esa huella a medida que avanzan.

Para ayudar, introducimos nuevos recursos para startups nativas de IA sobre cómo desplegar y servir modelos de Fireworks en Foundry y escalar desde prototipo hasta producción.

Plan de implementación diseñado para startups nativas de IA

El plan de implementación para desplegar modelos de IA de Fireworks muestra cómo los ingenieros fundadores y los pequeños equipos pueden pasar de la idea al MVP y al ajuste producto-mercado (PMF, por sus siglas en inglés) a través de la utilización de un enfoque repetible y nativo de Azure. La experiencia comienza de manera sencilla y crece según sus necesidades, así que ustedes son responsables de su inteligencia desde el principio.

La pila se ejecuta por completo dentro de su entorno Azure y solo requiere un endpoint modelo para la infraestructura o el harness de su aplicación. Empiecen con el despliegue de un único modelo, enrutar el tráfico a través de la Gestión de API, y seguir las métricas de latencia, uso y costes a lo largo del proceso. Cuando estén listos, pueden escalar con Azure Cache para Redis para reducir la inferencia redundante, a través de la introducción de ajustes de rendimiento según la carga de trabajo y desplegar múltiples variantes de modelos para pruebas A/B.

Los modelos de Fireworks se despliegan a través de Foundry dentro de su suscripción a Azure, por lo que el descubrimiento, la gobernanza y la facturación del modelo permanecen en un único plano de control.

Componente	Propósito
Microsoft Foundry + Modelos de IA Fireworks	Foundry proporciona la plataforma nativa de Azure para el despliegue, la gobernanza y la facturación; Fireworks son la capa de inferencia que sirve a los modelos abiertos que hay detrás de ellos
Azure Container Apps	Aloja la aplicación o API que envía solicitudes de inferencia
Azure Container Registry	Almacena imágenes de contenedores
Azure Key Vault	Almacena credenciales e información de endpoints de manera segura
Azure Monitor (opcional)	Proporciona observabilidad y conocimiento del rendimiento

Por qué las startups necesitan una arquitectura flexible de inferencia de IA

La inferencia es uno de los mayores factores de coste controlables para las empresas nativas de IA. Las decisiones tempranas sobre cómo se sirven los modelos pueden crear restricciones a largo plazo en costes, latencia y flexibilidad. Esta arquitectura está diseñada para abordar estos desafíos desde el principio. Servir modelos abiertos de esta manera mantiene esas decisiones en sus manos, para que puedan elegir, optimizar y cambiar los modelos detrás de su producto a medida que cambien sus necesidades de coste y rendimiento.

Optimizar el coste desde el primer día

• Utilicen inferencia serverless y de pago por token a través de Foundry con una selección de modelos abiertos
• Ajusten las cargas de trabajo al modelo más rentable y eviten estar atado a un solo proveedor de modelos
• Almacenar en caché las solicitudes repetidas con Azure Cache para Redis para reducir el uso de cómputo
• Rastrear el coste por millón de tokens como métrica central de ingeniería

Eliminar la sobrecarga de infraestructura

• No hace falta levantarse ni gestionar clústeres de GPU
• Fireworks proporciona inferencia de alto rendimiento, mientras que Foundry proporciona gobernanza, seguridad y gestión del ciclo de vida

Mantener la flexibilidad mientras escalas

• Experimentar y cambiar de modelo mediante APIs y flujos de despliegue consistentes, para que el intercambio requiera menos reestructuración
• Soportar pesos personalizados o para traer su propio modelo cuando sea necesario
• Pasar de la experimentación a la producción en la misma plataforma
• Una vez que la carga de trabajo está bien comprendida, sus suites de evaluación, bibliotecas de prompts y tráfico de producción graduado son datos de entrenamiento: los equipos pueden afinar y optimizar un modelo mediante Fireworks Training y luego importar a Azure al traer sus propios pesos.

Construir y probar aplicaciones de IA con menos presión inicial de costes

Para los equipos del programa Microsoft for Startups, esta arquitectura desbloquea una ventaja significativa. Pueden aplicar sus créditos de inicio a despliegues de modelos Fireworks a través de Data Zone Standard (las unidades de rendimiento provisionadas, o PTU, tienen capacidad reservada y no están cubiertas por créditos de startup), así como la infraestructura de Azure de apoyo.

Esto significa que pueden construir y probar aplicaciones de IA de calidad productiva, experimentar con múltiples modelos para encontrar dónde los modelos abiertos les dan la ventaja adecuada de coste y rendimiento antes de escalar, e iterar con rapidez hacia el ajuste producto-mercado, sin introducir presión inmediata sobre los costes de infraestructura.

Construyan con Microsoft para startups

Si están en el proceso de desarrollo de aplicaciones de IA en Azure, nos encantaría conocer más sobre su visión y ayudarles a acelerar su camino.

Microsoft for Startups ayuda a los fundadores a construir rápido, de manera inteligente y a vender más gracias a créditos de startup, infraestructura de IA Azure, orientación técnica y recursos de comercialización diseñados para ayudar a las startups a pasar de prototipos a despliegue empresarial más rápido. Empiecen hoy mismo con Microsoft for Startups.

Etiquetas:

• IA

The post Cómo desplegar Fireworks AI en Microsoft Foundry: un plan de arquitectura para startups appeared first on Source LATAM.

​The post Cómo desplegar Fireworks AI en Microsoft Foundry: un plan de arquitectura para startups appeared first on Source LATAM.  

Security teams can now access Daybreak Red and Daybreak Blue from OpenAI on Amazon Bedrock. Both are part of Daybreak, the cyber defense initiative from OpenAI that gives defenders governed access to frontier AI for cybersecurity work.

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. Both models run on Bedrock’s next-generation inference engine with zero-operator access (ZOA) enforced at the chip. Your inference data is not used for model training, and neither model requires you to opt into sharing your data with OpenAI.

Daybreak Red: GPT-5.6 Cyber and Daybreak Blue: GPT-5.6 Sol are now available to eligible customers on Amazon Bedrock in the following AWS Region: US East (N. Virginia). 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. To learn more, read the blog.

​Security teams can now access Daybreak Red and Daybreak Blue from OpenAI on Amazon Bedrock. Both are part of Daybreak, the cyber defense initiative from OpenAI that gives defenders governed access to frontier AI for cybersecurity work.
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. Both models run on Bedrock’s next-generation inference engine with zero-operator access (ZOA) enforced at the chip. Your inference data is not used for model training, and neither model requires you to opt into sharing your data with OpenAI.
Daybreak Red: GPT-5.6 Cyber and Daybreak Blue: GPT-5.6 Sol are now available to eligible customers on Amazon Bedrock in the following AWS Region: US East (N. Virginia). 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. To learn more, read the blog.  

Amazon S3 now includes the specific AWS Identity and Access Management (IAM) and AWS Organizations policy Amazon Resource Name (ARN) in HTTP 403 Access Denied error messages for same-account and same-organization requests. This helps you quickly identify the exact policy responsible for a denied request and remediate the issue directly.

Previously, S3 access denied error messages included the policy type and reason for denial, but when multiple policies of the same type existed, you still had to manually inspect each one to pinpoint the root cause. Now the error message includes the specific policy ARN for explicit deny cases, covering Service Control Policies (SCPs), Resource Control Policies (RCPs), identity-based policies, session policies, and permission boundaries.

This capability is available in all AWS Regions, including the AWS GovCloud (US) Regions and the AWS China Regions. To learn more about how to troubleshoot access denied errors in Amazon S3, visit the S3 User Guide and the IAM troubleshooting documentation.

​Amazon S3 now includes the specific AWS Identity and Access Management (IAM) and AWS Organizations policy Amazon Resource Name (ARN) in HTTP 403 Access Denied error messages for same-account and same-organization requests. This helps you quickly identify the exact policy responsible for a denied request and remediate the issue directly.
Previously, S3 access denied error messages included the policy type and reason for denial, but when multiple policies of the same type existed, you still had to manually inspect each one to pinpoint the root cause. Now the error message includes the specific policy ARN for explicit deny cases, covering Service Control Policies (SCPs), Resource Control Policies (RCPs), identity-based policies, session policies, and permission boundaries.
This capability is available in all AWS Regions, including the AWS GovCloud (US) Regions and the AWS China Regions. To learn more about how to troubleshoot access denied errors in Amazon S3, visit the S3 User Guide and the IAM troubleshooting documentation.  

Today, AWS announces the general availability of role manager, a capability in AWS Identity and Access Management (IAM) that automatically sets up the IAM roles your AWS services need. When you set up a supported service in the console, role manager creates a default role on your behalf, or reuses one that already exists in your account if it already matches the required permissions. You can enable or disable role manager at any time, as well as inspect the AWS-managed templates that role manager deploys on your behalf.

Role manager supports 6 AWS service consoles at launch, including AWS Lambda and Amazon EventBridge. For example, when you create an AWS Lambda function, role manager applies the AWS-managed template for that workflow. Roles created via role manager appear in the IAM console as standard IAM roles that you fully control, and you can identify the ones role manager created. When you are ready to tighten permissions, you can disable role manager and use IAM Access Analyzer to refine each role to only the permissions it needs.

Role manager is available in all AWS Regions, except the AWS GovCloud (US) Regions and the China Regions.

To learn more, see How AWS IAM role manager rethinks the starting point for IAM roles on the AWS Security Blog, or Create roles automatically with role manager in the IAM User Guide.

​Today, AWS announces the general availability of role manager, a capability in AWS Identity and Access Management (IAM) that automatically sets up the IAM roles your AWS services need. When you set up a supported service in the console, role manager creates a default role on your behalf, or reuses one that already exists in your account if it already matches the required permissions. You can enable or disable role manager at any time, as well as inspect the AWS-managed templates that role manager deploys on your behalf.
Role manager supports 6 AWS service consoles at launch, including AWS Lambda and Amazon EventBridge. For example, when you create an AWS Lambda function, role manager applies the AWS-managed template for that workflow. Roles created via role manager appear in the IAM console as standard IAM roles that you fully control, and you can identify the ones role manager created. When you are ready to tighten permissions, you can disable role manager and use IAM Access Analyzer to refine each role to only the permissions it needs.
Role manager is available in all AWS Regions, except the AWS GovCloud (US) Regions and the China Regions.
To learn more, see How AWS IAM role manager rethinks the starting point for IAM roles on the AWS Security Blog, or Create roles automatically with role manager in the IAM User Guide.  

Today, AWS announces the addition of an interactive map view to AWS Global View in the AWS Management Console, providing a visual way to explore AWS Global Infrastructure.

Previously, customers had to scan through a list of AWS locations in AWS Global View. With this new capability, customers can toggle between the interactive map view and the existing list view, making it easier to visualize their global AWS infrastructure footprint. 

When customers select the map view, they will see all AWS Regions and AWS Local Zones plotted on an interactive map. This capability helps customers make informed infrastructure planning decisions by visualizing already enabled AWS locations and the full range of AWS locations available to them, all in a single glance.

This capability is available across all public AWS Regions. To get started, navigate to the Regions and Zones page in AWS Global View console. For more information, see the AWS Global View documentation.

​Today, AWS announces the addition of an interactive map view to AWS Global View in the AWS Management Console, providing a visual way to explore AWS Global Infrastructure.
Previously, customers had to scan through a list of AWS locations in AWS Global View. With this new capability, customers can toggle between the interactive map view and the existing list view, making it easier to visualize their global AWS infrastructure footprint. 
When customers select the map view, they will see all AWS Regions and AWS Local Zones plotted on an interactive map. This capability helps customers make informed infrastructure planning decisions by visualizing already enabled AWS locations and the full range of AWS locations available to them, all in a single glance.
This capability is available across all public AWS Regions. To get started, navigate to the Regions and Zones page in AWS Global View console. For more information, see the AWS Global View documentation.  

Amazon Quick custom permissions now include deny by default, a governance setting that automatically restricts new AI capabilities before they reach users.

Previously, new AI capabilities were available to all users on release, requiring administrators to react after the fact. With deny by default, administrators restrict the AI capability category in a custom permissions profile and assign it to users, roles, or the entire account. Quick then denies any new AI capability at launch for those users. Restricting a category also restricts existing capabilities in it. Administrators explicitly allow each capability when ready. The restriction applies only to the profile you configure.

Configure deny by default in Manage account in Amazon Quick or through the AWS CLI. To learn more, see Custom permissions deny by default. Deny by default is available in all AWS Regions where Amazon Quick is available.

​Amazon Quick custom permissions now include deny by default, a governance setting that automatically restricts new AI capabilities before they reach users. Previously, new AI capabilities were available to all users on release, requiring administrators to react after the fact. With deny by default, administrators restrict the AI capability category in a custom permissions profile and assign it to users, roles, or the entire account. Quick then denies any new AI capability at launch for those users. Restricting a category also restricts existing capabilities in it. Administrators explicitly allow each capability when ready. The restriction applies only to the profile you configure. Configure deny by default in Manage account in Amazon Quick or through the AWS CLI. To learn more, see Custom permissions deny by default. Deny by default is available in all AWS Regions where Amazon Quick is available.