The week after AWS re:Invent builds on the excitement and energy of the event and is a good time to learn more and understand how the recent announcements can help you solve your challenges and unlock new opportunities. As usual, we have you covered with our top announcements of AWS re:Invent 2025 that you can learn all about here.

For me, one moment stood out above all the technical announcements: watching Rafi (Raphael Francis Quisumbing) from the Philippines receive the Now Go Build Award from Werner Vogels. Rafi has been an AWS Hero since 2015 and co-lead of AWS User Group Philippines since 2013. His dedication to building communities and empowering developers across the region embodies what this award represents. You can read more about Rafi on The Kernel. Congrats, Rafi!

The keynote recap: Agents, renaissance, and the developer’s role
This year’s AWS re:Invent keynotes painted a clear picture of where we’re headed.

Matt Garman emphasized that developers are “the heart of AWS” and that “freedom to invent” remains AWS’s core mission after 20 years. He focused on AI agents as the next inflection point: “AI assistants are starting to give way to AI agents that can perform tasks and automate on your behalf. This is where we’re starting to see material business returns from your AI investments.”

Swami Sivasubramanian highlighted the transformative moment we’re in: “For the first time in history, we can describe what we want to accomplish in natural language, and agents generate the plan. They write the code, call the necessary tools, and execute the complete solution.” AWS is building production-ready infrastructure that’s secure, reliable, and scalable—purpose-built for the non-deterministic nature of agents.

Peter DeSantis and Dave Brown reinforced that the core attributes AWS has obsessed over for 20 years—security, availability, performance, elasticity, cost, and agility—are more important than ever in the AI era. Dave Brown showcased Graviton and AWS’s custom silicon innovations that deliver these attributes at scale.

Werner Vogels delivered his final keynote after 14 years, introducing the concept of the “renaissance developer”—someone who is curious, thinks in systems, and communicates effectively. His message about AI and developer evolution resonated: “Will AI take my job? Maybe. Will AI make me obsolete? Absolutely not… if you evolve.” He emphasized that developers must be owners: “The work is yours, not that of the tools. You build it, you own it.”

You can also watch from keynotes, innovation talks to breakout sessions and more in the on-demand video page.

Innovations Talks

• Harnessing analytics for humans and AI (INV201)
• AI agents in action: Architecting the future of applications (INV202)
• The agent-enabled workplace: Transforming businesses with AI (INV203)
• Build and scale AI: from reliable agents to transformative systems (INV204)
• Reinventing software development with AI agents (INV205)
• Unlocking possibilities with AWS Compute (INV207)
• Databases made effortless so agents and developers can change the world (INV208)
• The next frontier: Building the agentic future of Financial Services (INV209)
• Infrastructure for the impossible: Turning public sector barriers into breakthroughs (INV210)
• Behind the curtain: How Amazon’s AI innovations are powered by AWS (INV211)
• Migrate, modernize, and move your business into the AI era (INV212)
• The power of cloud network innovation (INV213)
• Intelligent security: Protection at scale from development to production (INV214)
• AWS storage beyond data boundaries: Building the data foundation (INV215)

Breakout sessions — Topics	Breakout sessions — Segments
Analytics Application Integration Architecture Artificial Intelligence Business Applications Cloud Operations Compute Database Developer Tools End-User Computing Hybrid Cloud & Multi Cloud Industry Solutions Migration & Modernization Networking & Content Delivery Open Source Security & Identity Serverless & Containers Storage	Developer Community Digital Native Business Enterprise Independent Software Vendor New to AWS Partner Enablement Public Sector Senior Leaders Small & Medium Business Startup

Last week’s launches
Here are the launches that caught my attention not yet covered in our top announcements of AWS re:Invent 2025 post:

• Kiro Autonomous Agent – Building on Kiro’s general availability in November with team features, AWS introduced an autonomous agent that maintains awareness across sessions, learns from pull requests and feedback, and handles bug triage and code coverage improvements spanning multiple repositories. “Orders of magnitude more efficient” than first-generation AI coding tools, Matt Garman said. Kiro is now Amazon’s standard AI development environment company-wide.
• Multimodal Retrieval for Bedrock Knowledge Bases (GA) – Build AI-powered search and question-answering applications that work across text, images, audio, and video files. Developers can now ingest multimodal content with full control of parsing, chunking, embedding, and vector storage options, then send text or image queries to retrieve relevant segments across all media types.
• AWS Interconnect – Multicloud (Preview) – Quickly establish private, secure, high-speed network connections with dedicated bandwidth and built-in resiliency between Amazon VPCs and other cloud environments. Starting in preview with Google Cloud as the first launch partner, with Microsoft Azure support coming in 2026.

See AWS What’s New for more launch news that I haven’t covered here. That’s all for this week. Check back next Monday for another Weekly Roundup!

Happy building!

— Donnie

This post is part of our Weekly Roundup series. Check back each week for a quick roundup of interesting news and announcements from AWS!

Organizations face a challenging trade-off when adapting AI models to their specific business needs: settle for generic models that produce average results, or tackle the complexity and expense of advanced model customization. Traditional approaches force a choice between poor performance with smaller models or the high costs of deploying larger model variants and managing complex infrastructure. Reinforcement fine-tuning is an advanced technique that trains models using feedback instead of massive labeled datasets, but implementing it typically requires specialized ML expertise, complicated infrastructure, and significant investment—with no guarantee of achieving the accuracy needed for specific use cases.

Today, we’re announcing reinforcement fine-tuning in Amazon Bedrock, a new model customization capability that creates smarter, more cost-effective models that learn from feedback and deliver higher-quality outputs for specific business needs. Reinforcement fine-tuning uses a feedback-driven approach where models improve iteratively based on reward signals, delivering 66% accuracy gains on average over base models.

Amazon Bedrock automates the reinforcement fine-tuning workflow, making this advanced model customization technique accessible to everyday developers without requiring deep machine learning (ML) expertise or large labeled datasets.

How reinforcement fine-tuning works
Reinforcement fine-tuning is built on top of reinforcement learning principles to address a common challenge: getting models to consistently produce outputs that align with business requirements and user preferences.

While traditional fine-tuning requires large, labeled datasets and expensive human annotation, reinforcement fine-tuning takes a different approach. Instead of learning from fixed examples, it uses reward functions to evaluate and judge which responses are considered good for particular business use cases. This teaches models to understand what makes a quality response without requiring massive amounts of pre-labeled training data, making advanced model customization in Amazon Bedrock more accessible and cost-effective.

Here are the benefits of using reinforcement fine-tuning in Amazon Bedrock:

• Ease of use – Amazon Bedrock automates much of the complexity, making reinforcement fine-tuning more accessible to developers building AI applications. Models can be trained using existing API logs in Amazon Bedrock or by uploading datasets as training data, eliminating the need for labeled datasets or infrastructure setup.
• Better model performance – Reinforcement fine-tuning improves model accuracy by 66% on average over base models, enabling optimization for price and performance by training smaller, faster, and more efficient model variants. This works with Amazon Nova 2 Lite model, improving quality and price performance for specific business needs, with support for additional models coming soon.
• Security – Data remains within the secure AWS environment throughout the entire customization process, mitigating security and compliance concerns.

The capability supports two complementary approaches to provide flexibility for optimizing models:

• Reinforcement Learning with Verifiable Rewards (RLVR) uses rule-based graders for objective tasks like code generation or math reasoning.
• Reinforcement Learning from AI Feedback (RLAIF) employs AI-based judges for subjective tasks like instruction following or content moderation.

Getting started with reinforcement fine-tuning in Amazon Bedrock
Let’s walk through creating a reinforcement fine-tuning job.

First, I access the Amazon Bedrock console. Then, I navigate to the Custom models page. I choose Create and then choose Reinforcement fine-tuning job.

I start by entering the name of this customization job and then select my base model. At launch, reinforcement fine-tuning supports Amazon Nova 2 Lite, with support for additional models coming soon.

Next, I need to provide training data. I can use my stored invocation logs directly, eliminating the need to upload separate datasets. I can also upload new JSONL files or select existing datasets from Amazon Simple Storage Service (Amazon S3). Reinforcement fine-tuning automatically validates my training dataset and supports the OpenAI Chat Completions data format. If I provide invocation logs in the Amazon Bedrock invoke or converse format, Amazon Bedrock automatically converts them to the Chat Completions format.

The reward function setup is where I define what constitutes a good response. I have two options here. For objective tasks, I can select Custom code and write custom Python code that gets executed through AWS Lambda functions. For more subjective evaluations, I can select Model as judge to use foundation models (FMs) as judges by providing evaluation instructions.

Here, I select Custom code, and I create a new Lambda function or use an existing one as a reward function. I can start with one of the provided templates and customize it for my specific needs.

I can optionally modify default hyperparameters like learning rate, batch size, and epochs.

For enhanced security, I can configure virtual private cloud (VPC) settings and AWS Key Management Service (AWS KMS) encryption to meet my organization’s compliance requirements. Then, I choose Create to start the model customization job.

During the training process, I can monitor real-time metrics to understand how the model is learning. The training metrics dashboard shows key performance indicators including reward scores, loss curves, and accuracy improvements over time. These metrics help me understand whether the model is converging properly and if the reward function is effectively guiding the learning process.

When the reinforcement fine-tuning job is completed, I can see the final job status on the Model details page.

Once the job is completed, I can deploy the model with a single click. I select Set up inference, then choose Deploy for on-demand.

Here, I provide a few details for my model.

After deployment, I can quickly evaluate the model’s performance using the Amazon Bedrock playground. This helps me to test the fine-tuned model with sample prompts and compare its responses against the base model to validate the improvements. I select Test in playground.

The playground provides an intuitive interface for rapid testing and iteration, helping me confirm that the model meets my quality requirements before integrating it into production applications.

Interactive demo
Learn more by navigating an interactive demo of Amazon Bedrock reinforcement fine-tuning in action.

Additional things to know
Here are key points to note:

• Templates — There are seven ready-to-use reward function templates covering common use cases for both objective and subjective tasks.
• Pricing — To learn more about pricing, refer to the Amazon Bedrock pricing page.
• Security — Training data and custom models remain private and aren’t used to improve FMs for public use. It supports VPC and AWS KMS encryption for enhanced security.

Get started with reinforcement fine-tuning by visiting the reinforcement fine-tuning documentation and by accessing the Amazon Bedrock console.

Happy building!
— Donnie

Updated on March 3, 2026 – Policy in Amazon Bedrock AgentCore is now generally available.
Updated on March 31, 2026 – Amazon Bedrock AgentCore Evaluations is now generally available.

—
Today, we’re announcing new capabilities in Amazon Bedrock AgentCore to further remove barriers holding AI agents back from production. Organizations across industries are already building on AgentCore, the most advanced agentic platform to build, deploy, and operate highly capable agents securely at any scale. In just 5 months since preview, the AgentCore SDK has been downloaded over 2 million times. For example:

• PGA TOUR, a pioneer and innovation leader in sports has built a multi-agent content generation system to create articles for their digital platforms. The new solution, built on AgentCore, enables the PGA TOUR to provide comprehensive coverage for every player in the field, by increasing content writing speed by 1,000 percent while achieving a 95 percent reduction in costs.
• Independent software vendors (ISVs) like Workday are building the software of the future on AgentCore. AgentCore Code Interpreter provides Workday Planning Agent with secure data protection and essential features for financial data exploration. Users can analyze financial and operational data through natural language queries, making financial planning intuitive and self-driven. This capability reduces time spent on routine planning analysis by 30 percent, saving approximately 100 hours per month.
• Grupo Elfa, a Brazilian distributor and retailer, relies on AgentCore Observability for complete audit traceability and real-time metrics of their agents, transforming their reactive processes into proactive operations. Using this unified platform, their sales team can handle thousands of daily price quotes while the organization maintains full visibility of agent decisions, helping achieve 100 percent traceability of agent decisions and interactions, and reduced problem resolution time by 50 percent.

As organizations scale their agent deployments, they face challenges around implementing the right boundaries and quality checks to confidently deploy agents. The autonomy that makes agents powerful also makes them hard to confidently deploy at scale, as they might access sensitive data inappropriately, make unauthorized decisions, or take unexpected actions. Development teams must balance enabling agent autonomy while ensuring they operate within acceptable boundaries and with the quality you require to put them in front of customers and employees.

The new capabilities available today take the guesswork out of this process and help you build and deploy trusted AI agents with confidence:

• Policy in AgentCore (Preview) – Defines clear boundaries for agent actions by intercepting AgentCore Gateway tool calls before they run using policies with fine-grained permissions.
• AgentCore Evaluations (Preview) – Monitors the quality of your agents based on real-world behavior using built-in evaluators for dimensions such as correctness and helpfulness, plus custom evaluators for business-specific requirements.

We’re also introducing features that expand what agents can do:

• Episodic functionality in AgentCore Memory – A new long-term strategy that helps agents learn from experiences and adapt solutions across similar situations for improved consistency and performance in similar future tasks.
• Bidirectional streaming in AgentCore Runtime – Deploys voice agents where both users and agents can speak simultaneously following a natural conversation flow.

Policy in AgentCore for precise agent control
Policy gives you control over the actions agents can take and are applied outside of the agent’s reasoning loop, treating agents as autonomous actors whose decisions require verification before reaching tools, systems, or data. It integrates with AgentCore Gateway to intercept tool calls as they happen, processing requests while maintaining operational speed, so workflows remain fast and responsive.

You can create policies using natural language or directly use Cedar—an open source policy language for fine-grained permissions—simplifying the process to set up, understand, and audit rules without writing custom code. This approach makes policy creation accessible to development, security, and compliance teams who can create, understand, and audit rules without specialized coding knowledge.

The policies operate independently of how the agent was built or which model it uses. You can define which tools and data agents can access—whether they are APIs, AWS Lambda functions, Model Context Protocol (MCP) servers, or third-party services—what actions they can perform, and under what conditions.

Teams can define clear policies once and apply them consistently across their organization. With policies in place, developers gain the freedom to create innovative agentic experiences, and organizations can deploy their agents to act autonomously while knowing they’ll stay within defined boundaries and compliance requirements.

Using Policy in AgentCore
You can start by creating a policy engine in the new Policy section of the AgentCore console and associate it with one or more AgentCore gateways.

A policy engine is a collection of policies that are evaluated at the gateway endpoint. When associating a gateway with a policy engine, you can choose whether to enforce the result of the policy—effectively permitting or denying access to a tool call—or to only emit logs. Using logs helps you test and validate a policy before enabling it in production.

Then, you can define the policies to apply to have granular control over access to the tools offered by the associated AgentCore gateways.

To create a policy, you can start with a natural language description (that should include information of the authentication claims to use) or directly edit Cedar code.

Natural language-based policy authoring provides a more accessible way for you to create fine-grained policies. Instead of writing formal policy code, you can describe rules in plain English. The system interprets your intent, generates candidate policies, validates them against the tool schema, and uses automated reasoning to check safety conditions—identifying prompts that are overly permissive, overly restrictive, or contain conditions that can never be satisfied.

Unlike generic large language model (LLM) translations, this feature understands the structure of your tools and generates policies that are both syntactically correct and semantically aligned with your intent, while flagging rules that cannot be enforced. It is also available as a Model Context Protocol (MCP) server, so you can author and validate policies directly in your preferred AI-assisted coding environment as part of your normal development workflow. This approach reduces onboarding time and helps you write high-quality authorization rules without needing Cedar expertise.

The following sample policy uses information from the OAuth claims in the JWT token used to authenticate to an AgentCore gateway (for the role) and the arguments passed to the tool call (context.input) to validate access to the tool processing a refund. Only an authenticated user with the refund-agent role can access the tool but for amounts (context.input.amount) lower than $200 USD.

permit(
  principal is AgentCore::OAuthUser,
  action == AgentCore::Action::"RefundTool__process_refund",
  resource == AgentCore::Gateway::"<GATEWAY_ARN>"
)
when {
  principal.hasTag("role") &&
  principal.getTag("role") == "refund-agent" &&
  context.input.amount < 200
};

AgentCore Evaluations for continuous, real-time quality intelligence
AgentCore Evaluations is a fully managed service that helps you continuously monitor and analyze agent performance based on real-world behavior. With AgentCore Evaluations, you can use built-in evaluators for common quality dimensions such as correctness, helpfulness, tool selection accuracy, safety, goal success rate, and context relevance. You can also create custom model-based scoring systems configured with your choice of prompt and model for business-tailored scoring while the service samples live agent interactions and scores them continuously.

All results from AgentCore Evaluations are visualized in Amazon CloudWatch alongside AgentCore Observability insights, providing one place for unified monitoring. You can also set up alerts and alarms on the evaluation scores to proactively monitor agent quality and respond when metrics fall outside acceptable thresholds.

You can use AgentCore Evaluations during the testing phase where you can check an agent against the baseline before deployment to stop faulty versions from reaching users, and in production for continuous improvement of your agents. When quality metrics drop below defined thresholds—such as a customer service agent satisfaction declining or politeness scores dropping by more than 10 percent over an 8-hour period—the system triggers immediate alerts, helping to detect and address quality issues faster.

Using AgentCore Evaluations
You can create an online evaluation in the new Evaluations section of the AgentCore console. You can use as data source an AgentCore agent endpoint or a CloudWatch log group used by an external agent. For example, I use here the same sample customer support agent I shared when we introduced AgentCore in preview.

Then, you can select the evaluators to use, including custom evaluators that you can define starting from the existing templates or build from scratch.

For example, for a customer support agent, you can select metrics such as:

• Correctness – Evaluates whether the information in the agent’s response is factually accurate
• Faithfulness – Evaluates whether information in the response is supported by provided context/sources
• Helpfulness – Evaluates from user’s perspective how useful and valuable the agent’s response is
• Harmfulness – Evaluates whether the response contains harmful content
• Stereotyping – Detects content that makes generalizations about individuals or groups

The evaluators for tool selection and tool parameter accuracy can help you understand if an agent is choosing the right tool for a task and extracting the correct parameters from the user queries.

To complete the creation of the evaluation, you can choose the sampling rate and optional filters. For permissions, you can create a new AWS Identity and Access Management (IAM) service role or pass an existing one.

The results are published, as they are evaluated, on Amazon CloudWatch in the AgentCore Observability dashboard. You can choose any of the bar chart sections to see the corresponding traces and gain deeper insight into the requests and responses behind that specific evaluation.

Because the results are in CloudWatch, you can use all of its feature to create, for example, alarms and automations.

Creating custom evaluators in AgentCore Evaluations
Custom evaluators allow you to define business-specific quality metrics tailored to your agent’s unique requirements. To create a custom evaluator, you provide the model to use as a judge, including inference parameters such as temperature and max output tokens, and a tailored prompt with the judging instructions. You can start from the prompt used by one of the built-in evaluators or enter a new one.

Then, you define the scale to produce in output. It can be either numeric values or custom text labels that you define. Finally, you configure whether the evaluation is computed by the model on single traces, full sessions, or for each tool call.

AgentCore Memory episodic functionality for experience-based learning
AgentCore Memory, a fully managed service that gives AI agents the ability to remember past interactions, now includes a new long-term memory strategy that gives agents the ability to learn from past experiences and apply those lessons to provide more helpful assistance in future interactions.

Consider booking travel with an agent: over time, the agent learns from your booking patterns—such as the fact that you often need to move flights to later times when traveling for work due to client meetings. When you start your next booking involving client meetings, the agent proactively suggests flexible return options based on these learned patterns. Just like an experienced assistant who learns your specific travel habits, agents with episodic memory can now recognize and adapt to your individual needs.

When you enable the new episodic functionality, AgentCore Memory captures structured episodes that record the context, reasoning process, actions taken, and outcomes of agent interactions, while a reflection agent analyzes these episodes to extract broader insights and patterns. When facing similar tasks, agents can retrieve these learnings to improve decision-making consistency and reduce processing time. This reduces the need for custom instructions by including in the agent context only the specific learnings an agent needs to complete a task instead of a long list of all possible suggestions.

AgentCore Runtime bidirectional streaming for more natural conversations
With AgentCore Runtime, you can deploy agentic applications with few lines of code. To simplify deploying conversational experiences that feel natural and responsive, AgentCore Runtime now supports bidirectional streaming. This capability enables voice agents to listen and adapt while users speak, so that people can interrupt agents mid-response and have the agent immediately adjust to the new context—without waiting for the agent to finish its current output. Rather than traditional turn-based interaction where users must wait for complete responses, bidirectional streaming creates flowing, natural conversations where agents dynamically change their response based on what the user is saying.

Building these conversational experiences from the ground up requires significant engineering effort to handle the complex flow of simultaneous communication. Bidirectional streaming simplifies this by managing the infrastructure needed for agents to process input while generating output, handling interruptions gracefully, and maintaining context throughout dynamic conversation shifts. You can now deploy agents that naturally adapt to the fluid nature of human conversation—supporting mid-thought interruptions, context switches, and clarifications without losing the thread of the interaction.

Things to know
Amazon Bedrock AgentCore, including the preview of Policy, is available in the US East (N. Virginia), US West (Oregon), Asia Pacific (Mumbai, Singapore, Sydney, Tokyo), and Europe (Frankfurt, Ireland) AWS Regions . The preview of AgentCore Evaluations is available in the US East (N. Virginia), US West (Oregon), Asia Pacific (Sydney), and Europe (Frankfurt) Regions. For Regional availability and future roadmap, visit AWS Capabilities by Region.

With AgentCore, you pay for what you use with no upfront commitments. For detailed pricing information, visit the Amazon Bedrock pricing page. AgentCore is also a part of the AWS Free Tier that new AWS customers can use to get started at no cost and explore key AWS services.

These new features work with any open source framework such as CrewAI, LangGraph, LlamaIndex, and Strands Agents, and with any foundation model. AgentCore services can be used together or independently, and you can get started using your favorite AI-assisted development environment with the AgentCore open source MCP server.

To learn more and get started quickly, visit the AgentCore Developer Guide.

— Danilo

Editors Note: 1/16/25- AgentCore Evaluation is not available in US East (Ohio). Removed from region list.

Today, we’re announcing the general availability of an additional 18 fully managed open weight models in Amazon Bedrock from Google, MiniMax AI, Mistral AI, Moonshot AI, NVIDIA, OpenAI, and Qwen, including the new Mistral Large 3 and Ministral 3 3B, 8B, and 14B models.

With this launch, Amazon Bedrock now provides nearly 100 serverless models, offering a broad and deep range of models from leading AI companies, so customers can choose the precise capabilities that best serve their unique needs. By closely monitoring both customer needs and technological advancements, we regularly expand our curated selection of models based on customer needs and technological advancements to include promising new models alongside established industry favorites.

This ongoing expansion of high-performing and differentiated model offerings helps customers stay at the forefront of AI innovation. You can access these models on Amazon Bedrock through the unified API, evaluate, switch, and adopt new models without rewriting applications or changing infrastructure.

New Mistral AI models
These four Mistral AI models are now available first on Amazon Bedrock, each optimized for different performance and cost requirements:

• Mistral Large 3 – This open weight model is optimized for long-context, multimodal, and instruction reliability. It excels in long document understanding, agentic and tool use workflows, enterprise knowledge work, coding assistance, advanced workloads such as math and coding tasks, multilingual analysis and processing, and multimodal reasoning with vision.
• Ministral 3 3B – The smallest in the Ministral 3 family is edge-optimized for single GPU deployment with strong language and vision capabilities. It shows robust performance in image captioning, text classification, real-time translation, data extraction, short content generation, and lightweight real-time applications on edge or low-resource devices.
• Ministral 3 8B – The best-in-class Ministral 3 model for text and vision is edge-optimized for single GPU deployment with high performance and minimal footprint. This model is ideal for chat interfaces in constrained environments, image and document description and understanding, specialized agentic use cases, and balanced performance for local or embedded systems.
• Ministral 3 14B – The most capable Ministral 3 model delivers state-of the-art text and vision performance optimized for single GPU deployment. You can use advanced local agentic use cases and private AI deployments where advanced capabilities meet practical hardware constraints.

More open weight model options
You can use these open weight models for a wide range of use cases across industries:

Model provider	Model name	Description	Use cases
Google	Gemma 3 4B	Efficient text and image model that runs locally on laptops. Multilingual support for on-device AI applications.	On-device AI for mobile and edge applications, privacy-sensitive local inference, multilingual chat assistants, image captioning and description, and lightweight content generation.
	Gemma 3 12B	Balanced text and image model for workstations. Multi-language understanding with local deployment for privacy-sensitive applications.	Workstation-based AI applications; local deployment for enterprises; multilingual document processing, image analysis and Q&A; and privacy-compliant AI assistants.
	Gemma 3 27B	Powerful text and image model for enterprise applications. Multi-language support with local deployment for privacy and control.	Enterprise local deployment, high-performance multimodal applications, advanced image understanding, multilingual customer service, and data-sensitive AI workflows.
Moonshot AI	Kimi K2 Thinking	Deep reasoning model that thinks while using tools. Handles research, coding and complex workflows requiring hundreds of sequential actions.	Complex coding projects requiring planning, multistep workflows, data analysis and computation, and long-form content creation with research.
MiniMax AI	MiniMax M2	Built for coding agents and automation. Excels at multi-file edits, terminal operations and executing long tool-calling chains efficiently.	Coding agents and integrated development environment (IDE) integration, multi-file code editing, terminal automation and DevOps, long-chain tool orchestration, and agentic software development.
Mistral AI	Magistral Small 1.2	Excels at math, coding, multilingual tasks, and multimodal reasoning with vision capabilities for efficient local deployment.	Math and coding tasks, multilingual analysis and processing, and multimodal reasoning with vision.
	Voxtral Mini 1.0	Advanced audio understanding model with transcription, multilingual support, Q&A, and summarization.	Voice-controlled applications, fast speech-to-text conversion, and offline voice assistants.
	Voxtral Small 1.0	Features state-of-the-art audio input with best-in-class text performance; excels at speech transcription, translation, and understanding.	Enterprise speech transcription, multilingual customer service, and audio content summarization.
NVIDIA	NVIDIA Nemotron Nano 2 9B	High efficiency LLM with hybrid transformer Mamba design, excelling in reasoning and agentic tasks.	Reasoning, tool calling, math, coding, and instruction following.
	NVIDIA Nemotron Nano 2 VL 12B	Advanced multimodal reasoning model for video understanding and document intelligence, powering Retrieval-Augmented Generation (RAG) and multimodal agentic applications.	Multi-image and video understanding, visual Q&A, and summarization.
OpenAI	gpt-oss-safeguard-20b	Content safety model that applies your custom policies. Classifies harmful content with explanations for trust and safety workflows.	Content moderation and safety classification, custom policy enforcement, user-generated content filtering, trust and safety workflows, and automated content triage.
	gpt-oss-safeguard-120b	Larger content safety model for complex moderation. Applies custom policies with detailed reasoning for enterprise trust and safety teams.	Enterprise content moderation at scale, complex policy interpretation, multilayered safety classification, regulatory compliance checking, high-stakes content review.
Qwen	Qwen3-Next-80B-A3B	Fast inference with hybrid attention for ultra-long documents. Optimized for RAG pipelines, tool use & agentic workflows with quick responses.	RAG pipelines with long documents, agentic workflows with tool calling, code generation and software development, multi-turn conversations with extended context, multilingual content generation.
	Qwen3-VL-235B-A22B	Understands images and video. Extracts text from documents, converts screenshots to working code, and automates clicking through interfaces.	Extracting text from images and PDFs, converting UI designs or screenshots to working code, automating clicks and navigation in applications, video analysis and understanding, reading charts and diagrams.

When implementing publicly available models, give careful consideration to data privacy requirements in your production environments, check for bias in output, and monitor your results for data security, responsible AI, and model evaluation.

You can access the enterprise-grade security features of Amazon Bedrock and implement safeguards customized to your application requirements and responsible AI policies with Amazon Bedrock Guardrails. You can also evaluate and compare models to identify the optimal models for your use cases by using Amazon Bedrock model evaluation tools.

To get started, you can quickly test these models with a few prompts in the playground of the Amazon Bedrock console or use any AWS SDKs to include access to the Bedrock InvokeModel and Converse APIs. You can also use these models with any agentic framework that supports Amazon Bedrock and deploy the agents using Amazon Bedrock AgentCore and Strands Agents. To learn more, visit Code examples for Amazon Bedrock using AWS SDKs in the Amazon Bedrock User Guide.

Now available
Check the full Region list for availability and future updates of new models or search your model name in the AWS CloudFormation resources tab of AWS Capabilities by Region. To learn more, check out the Amazon Bedrock product page and the Amazon Bedrock pricing page.

Give these models a try in the Amazon Bedrock console today and send feedback to AWS re:Post for Amazon Bedrock or through your usual AWS Support contacts.

— Channy

Updated on 4 December — Amazon Bedrock now supports Responses API on new OpenAI API-compatible service endpoints for GPT OSS 20B and 120B models. To learn more, visit Generate responses using OpenAI APIs.