Change
4ba02f4
4ba02f425d2644e924e0ed10ce8272e631205aba · commit on GitHub
aws-blog-feed: changed (536446 bytes, HTTP 200)
raw/aws-blog-feed/response.xml modified
- Source
- aws-blog-feed
- Lines added
- +3,245
- Lines removed
- -4,503
- Stored bytes at this commit
- 536,446
- Timestamp
- observed
- Raw artifact at this commit
- raw/aws-blog-feed/response.xml
Recorded headers
| observed_at | 2026-09-23T04:45:34.135Z |
|---|---|
| origin_date | null |
| status | 200 |
| final URL | https://aws.amazon.com/blogs/machine-learning/feed/ |
| etag | null |
| last-modified | Tue, 22 Sep 2026 22:38:50 GMT |
| date | Wed, 23 Sep 2026 04:45:34 GMT |
| age | null |
| cache-control | null |
| cf-cache-status | null |
| content-encoding | null |
| content-length | null |
@
@@ -5,7 +5,7 @@ <atom:link href="https://aws.amazon.com/blogs/machine-learning/feed/" rel="self" type="application/rss+xml"/> <link>https://aws.amazon.com/blogs/machine-learning/</link> <description>Official Machine Learning Blog of Amazon Web Services</description>-
<lastBuildDate>Mon, 21 Sep 2026 18:30:34 +0000</lastBuildDate>+
<lastBuildDate>Tue, 22 Sep 2026 18:10:22 +0000</lastBuildDate> <language>en-US</language> <sy:updatePeriod> hourly </sy:updatePeriod>@
@@ -13,291 +13,684 @@ 1 </sy:updateFrequency> <item>-
<title>xAI’s Grok 4.6 is now available in Amazon Bedrock</title>-
<link>https://aws.amazon.com/blogs/machine-learning/xais-grok-4-6-is-now-available-in-amazon-bedrock/</link>+
<title>Bring more intelligence to everyday work with GPT-6 Sol and GPT-6 Luna on Amazon Bedrock</title>+
<link>https://aws.amazon.com/blogs/machine-learning/bring-more-intelligence-to-everyday-work-with-gpt-6-sol-and-gpt-6-luna-on-amazon-bedrock/</link> -
<dc:creator><![CDATA[Suheel Farooq]]></dc:creator>-
<pubDate>Mon, 21 Sep 2026 18:30:34 +0000</pubDate>+
<dc:creator><![CDATA[Tanvi Girinath]]></dc:creator>+
<pubDate>Tue, 22 Sep 2026 18:10:22 +0000</pubDate> <category><![CDATA[Amazon Bedrock]]></category>-
<category><![CDATA[Amazon Machine Learning]]></category> <category><![CDATA[Announcements]]></category>-
<category><![CDATA[Artificial Intelligence]]></category>-
<category><![CDATA[Foundation models]]></category>-
<category><![CDATA[Generative AI]]></category> <category><![CDATA[Intermediate (200)]]></category>-
<category><![CDATA[Launch]]></category>-
<guid isPermaLink="false">56dfeac3bfef33ff7b19f174cf3e455627c66990</guid>+
<guid isPermaLink="false">f91c4872e4efa50ad6fa1a0161799f318feef3c1</guid>+
+
<description>GPT-6 Sol and GPT-6 Luna are now generally available on Amazon Bedrock, giving you more options to match intelligence and efficiency to each workload.</description>+
<content:encoded><p><em>GPT-6 Sol and GPT-6 Luna are now generally available on Amazon Bedrock, giving you more options to match intelligence and efficiency to each workload.</em></p>+
<p>The value of AI at scale depends on two dimensions: what a model can do and how often you can put it to use. Greater intelligence expands the complexity a model can handle, from subtle coding problems to multistep processes across tools. Efficiency determines how broadly that intelligence c…+
<p><a href="https://aws.amazon.com/blogs/machine-learning/take-on-your-most-ambitious-work-with-gpt-6-astra-on-amazon-bedrock/" target="_blank" rel="noopener">GPT-6 Astra</a> established the upper end of the GPT-6 family for the most ambitious projects, where achieving the highest-…+
<p>Today, <a href="https://aws.amazon.com/bedrock/openai/" target="_blank" rel="noopener">GPT-6 Sol and GPT-6 Luna</a> from OpenAI are generally available on <a href="https://aws.amazon.com/bedrock/" target="_blank" rel="noopener">Amazon Bedrock</a>, running on an infer…+
<h2 id="solve-harder-problems-every-day">Solve harder problems every day</h2>+
<p>GPT-6 Sol is designed for demanding tasks that recur throughout development and operations. It can implement features, debug issues, refactor and review code, analyze data, and complete multistep processes across tools and applications. Improvements over GPT-5.6 Sol in coding and computer u…+
<p>As GPT-6 Sol handles more of that process, developers need to see what it changed, what it verified, and what it could not confirm. On an internal factuality evaluation, OpenAI found that GPT-6 Sol made approximately half as many factual mistakes as GPT-5.6 Sol. GPT-6 Sol also benefits from…+
<p>Together, stronger execution and clearer reporting make GPT-6 Sol practical across the development cycle. The relevant measure there is the total cost of reaching a usable result, including output quality, token usage, retries, and latency.</p>+
<h2 id="make-focused-intelligence-economical-at-volume">Make focused intelligence economical at volume</h2>+
<p>When a task runs thousands of times a day, the economics of each call determine whether the workflow scales. A single classification or summary is inexpensive on its own, but the cost of extraction, routing, and follow-up across a full document pipeline compounds with every additional reque…+
<p>GPT-6 Luna is designed for workloads where that volume matters. You can use it to extract information from large document collections, summarize incoming material, classify inputs, and answer focused questions across many users or applications.</p>+
<p>Efficiency at volume also requires consistent&nbsp;outputs. OpenAI’s evaluations show improvements in GPT-6 Luna’s factual reliability and clearer communication of results. You can also adjust reasoning effort per request to balance the quality, responsiveness, and cost each task requir…+
<h2 id="match-intelligence-to-each-step-without-rebuilding-context">Match intelligence to each step without rebuilding context</h2>+
<p>A single application may need different levels of intelligence as a request progresses. You might use GPT-6 Luna to classify incoming requests, GPT-6 Sol to investigate complex cases, and GPT-6 Astra when additional reasoning depth can materially change a decision. This concentrates intelli…+
<p>Within each stage, repeated calls to the same model may reuse instructions, tool definitions, policies, and reference material. Reprocessing that context can erode the efficiency gained by selecting the appropriate model.</p>+
<p>GPT-6 Sol and GPT-6 Luna support explicit prompt caching on Amazon Bedrock. You can mark prompt content for reuse, allowing subsequent requests to focus processing on new input. This is useful for coding assistants that reuse repository instructions, support applications grounded in the sam…+
<h2 id="run-gpt-6-at-scale-with-performance-and-control">Run GPT-6 at scale with performance and control</h2>+
<p>As AI usage grows, model quality is only part of what determines whether an application succeeds in production. Teams also need infrastructure that maintains performance as demand changes, economics that hold across repeated requests, and controls that protect sensitive data. Amazon Bedrock…+
<p>You can govern model access through AWS Identity and Access Management (IAM) policies and audit every invocation through AWS CloudTrail. Virtual private cloud (VPC) endpoints powered by AWS PrivateLink help keep traffic within your network boundaries. Inference runs on hardware-isolated inf…+
<p>Your inference data isn’t used for model training, and using GPT-6 Sol and GPT-6 Luna doesn’t require you to opt into sharing your data with OpenAI. For <a href="https://docs.aws.amazon.com/bedrock/latest/userguide/abuse-detection.html" target="_blank" rel="noopener">automated abuse d…+
<h2 id="get-started">Get started</h2>+
<p>You can get started with GPT-6 Sol and GPT-6 Luna in the <a href="https://us-east-1.console.aws.amazon.com/bedrock/home?region=us-east-1#/" target="_blank" rel="noopener">Amazon Bedrock console</a> or programmatically through supported Amazon Bedrock APIs. For information about …+
<p><em>Interested in how Amazon Bedrock can support your team?</em> <em><a href="https://pages.awscloud.com/Amazon-Bedrock-Contact-Us.html" target="_blank" rel="noopener">Connect with us</a> to start the conversation.</em></p>+
<hr style="width: 100%">+
<h2>About the authors</h2>+
<footer>+
<div class="blog-author-box" style="padding-top: 2.0em">+
<div class="blog-author-image" style="margin-right: 1.0em">+
<p><img class="alignnone size-full wp-image-139734" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/09/22/ML-21636-6.jpg" alt="" width="300" height="400"></p>+
</div>+
<h3 class="lb-h4">Tanvi Girinath</h3>+
<p>Tanvi is a Product Marketing Manager for Amazon Bedrock at Amazon Web Services (AWS), where she helps customers adopt and scale AI applications and agents with Amazon Bedrock.</p>+
</div>+
<div class="blog-author-box" style="padding-top: 2.0em">+
<div class="blog-author-image" style="margin-right: 1.0em">+
<p><img loading="lazy" class="alignnone size-full" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/09/21/ML-21956-2.jpg" alt="Chris Dickens" width="100" height="133"></p>+
</div>+
<h3 class="lb-h4">Chris Dickens</h3>+
<p>Chris is a Member of Product Staff at OpenAI focused on the OpenAI APIs. His work includes collaboration with AWS on Amazon Bedrock to make OpenAI’s frontier models widely accessible to developers.</p>+
</div>+
<div class="blog-author-box" style="padding-top: 2.0em">+
<div class="blog-author-image" style="margin-right: 1.0em">+
<img loading="lazy" class="alignnone size-full" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/09/03/ML-21849-3.jpg" alt="Manish Rathaur" width="100" height="133">+
</div>+
<h3 class="lb-h4">Manish Rathaur</h3>+
<p style="overflow: hidden">Manish is a Senior Product Manager for Amazon Bedrock.</p>+
</div>+
</footer></content:encoded>+
+
+
+
</item>+
<item>+
<title>Claude Opus 5.5 is now available on AWS</title>+
<link>https://aws.amazon.com/blogs/machine-learning/claude-opus-5-5-is-now-available-on-aws/</link>+
+
<dc:creator><![CDATA[Aamna Najmi]]></dc:creator>+
<pubDate>Tue, 22 Sep 2026 17:28:01 +0000</pubDate>+
<category><![CDATA[Amazon Bedrock]]></category>+
<category><![CDATA[Announcements]]></category>+
<category><![CDATA[Intermediate (200)]]></category>+
<guid isPermaLink="false">2e3d91b641ca237ade2f2527a383a24847fe5ded</guid>+
+
<description>Claude Opus 5.5, Anthropic's most capable Opus model for agentic coding, knowledge work, and long-running tasks, is now available on Amazon Bedrock and Claude Platform on AWS. This post covers what's new in Opus 5.5, practical guidance, and how to start building with the model on A…+
<content:encoded><p>Today, we’re excited to announce the availability of Claude Opus 5.5 on <a href="https://aws.amazon.com/bedrock/" target="_blank" rel="noopener">Amazon Bedrock</a> and <a href="https://aws.amazon.com/blogs/machine-learning/introducing-claude-platfor…+
<p>This post covers Claude Opus 5.5’s improvements, practical guidance, and how to start building with the model on Amazon Bedrock.</p>+
<h2 id="what-makes-claude-opus-5.5-different">What makes Claude Opus 5.5 different</h2>+
<p>According to Anthropic, Claude Opus 5.5 does more with fewer tokens than Claude Opus 5, and new pricing passes those gains straight to customers. Lower per-token prices and much cheaper cache reads stack on top of the efficiency gains. The result is an average lower cost per task than Claud…+
<p>Claude Opus 5.5 is trained to communicate more clearly. As it works, it surfaces what it did, what it found, and what it needs, making long-running tasks easier to follow. Adaptive thinking is always on, and Opus 5.5 decides how much reasoning each task needs. You can use effort as your con…+
<p>Claude Opus 5.5 is the first Opus model that comes with safety classifiers similar to Claude Fable 5.1 in biology, cyber security, and AI development. Requests will be refused more frequently as compared to previous Opus versions.</p>+
<h2 id="use-cases">Use cases</h2>+
<p>Claude Opus 5.5 capabilities are a good fit for industries where consistency and depth matter most. In software development, Opus 5.5 is an improvement over Opus 5 for longer-running sessions with clear communication and explainability, making it easier to use, review, and trust. For knowle…+
<h2 id="getting-started-with-claude-opus-5.5-on-amazon-bedrock">Getting started with Claude Opus 5.5 on Amazon Bedrock</h2>+
<p>To try Claude Opus 5.5, open the <a href="https://console.aws.amazon.com/bedrock/" target="_blank" rel="noopener">Amazon Bedrock console</a>, choose <strong>Test</strong>, then <strong>Playground</strong>, and select <strong>Claude Opus 5.5</stro…+
<div id="attachment_140005" style="width: 1876px" class="wp-caption alignnone">+
<img aria-describedby="caption-attachment-140005" loading="lazy" class="size-full wp-image-140005" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/09/22/Screenshot-2026-09-22-at-18.37.23.png" alt="" width="1866" height="986">+
<p id="caption-attachment-140005" class="wp-caption-text">Figure 1: Selecting an Anthropic Claude model in the Amazon Bedrock console Playground</p>+
</div>+
<div class="mceTemp">+
<div id="attachment_140006" style="width: 1627px" class="wp-caption alignnone">+
<img aria-describedby="caption-attachment-140006" loading="lazy" class="size-full wp-image-140006" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/09/22/Screenshot-2026-09-22-at-18.36.48.png" alt="" width="1617" height="906">+
<p id="caption-attachment-140006" class="wp-caption-text">Figure 2: Running a prompt against a Claude model in the Amazon Bedrock console Playground</p>+
</div>+
<p>Programmatically, you can call the model with the <a href="https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-anthropic-claude-messages.html" target="_blank" rel="noopener">Anthropic Messages API</a> against <code>bedrock-runtime</code> and <co…+
<h3 id="prerequisites">Prerequisites</h3>+
<ol type="1">+
<li>Active AWS account with Amazon Bedrock access.</li>+
<li>AWS Command Line Interface (AWS CLI) installed and configured.</li>+
<li>Python 3.10+.</li>+
<li>Boto3 installed: <code>pip install boto3</code>.</li>+
<li>Anthropic SDK installed: <code>pip install anthropic</code>.</li>+
<li>The Amazon Bedrock Token Generator for Amazon Bedrock authentication installed: <code>pip install aws_bedrock_token_generator</code>.</li>+
<li>AWS Identity and Access Management (IAM) permissions: <code>bedrock:InvokeModel</code> and <code>bedrock:InvokeModelWithResponseStream</code>.</li>+
</ol>+
<p>Here’s a quick example using the AWS SDK for Python (Boto3):</p>+
<div class="hide-language">+
<pre><code class="language-python">import boto3+
import json+
+
# Create a Bedrock Runtime client+
bedrock_runtime = boto3.client(+
service_name="bedrock-runtime",+
region_name="us-east-1"+
)-
<description>xAI's Grok 4.6 is now available in Amazon Bedrock: a frontier model for long-running agents, coding, and knowledge work, with a 500K token context window and four reasoning effort levels. It runs on both the bedrock-mantle and bedrock-runtime endpoints, with Converse API and cross-…-
<content:encoded><p>Today, we are announcing that xAI’s Grok 4.6 is available in <a href="https://aws.amazon.com/bedrock/" target="_blank" rel="noopener">Amazon Bedrock</a>, adding a frontier model built for long-running agents, coding, and knowledge work to the Bedrock m…-
<p>This is xAI’s second model in Amazon Bedrock. When <a href="https://aws.amazon.com/blogs/machine-learning/introducing-grok-on-amazon-bedrock/" target="_blank" rel="noopener">Grok 4.3 became generally available</a>, xAI joined Amazon Bedrock as a model provider and the model was …-
<p>This post covers what xAI says Grok 4.6 is designed for, how it is packaged on Amazon Bedrock, and how to send your first request.</p>-
<h2 id="what-grok-4.6-is-built-for">What Grok 4.6 is built for</h2>-
<p>The capability and training details in this section come from xAI’s launch announcement, <a href="https://x.ai/news/grok-4-6" target="_blank" rel="noopener">Introducing Grok 4.6</a>.</p>-
<p>Grok 4.6 builds on Grok 4.5 with a particular focus on long-running agents and more ambitious interactive and visual work. xAI describes the model as staying with complex tasks across many steps, whether that is researching a topic, analyzing information, working across a code base, or turn…-
<p>On training, xAI reports a longer supplemental training run than Grok 4.5, using curated model-generated data for reasoning and advanced technical concepts, high-quality engineering data, and an improved optimizer and training recipe. It then used Grok 4.5 to regenerate the supervised fine-…-
<p>Two behaviors xAI calls out are worth noting for anyone building agents. On longer trajectories, the model began showing more self-testing and verification, checking its own work before moving on. It also produces stronger first passes on visual and interactive projects, establishing the st…-
<p>On safety, xAI states that Grok 4.6’s safeguards have been improved and calibrated in line with the model’s capabilities, backed by what it describes as its widest-ever suite of pre-deployment testing for capabilities and safeguard calibration, plus post-deployment and third-party testing. …-
<h2 id="reported-benchmark-results">Reported benchmark results</h2>-
<p>xAI reports that Grok 4.6 achieves frontier intelligence across several agentic coding and knowledge work benchmarks. These are the figures it published for Grok 4.6 High at launch on August 12, 2026:</p>-
<table style="height: 490px" border="1px" width="856" cellpadding="10px">+
# Invoke Claude Opus 5.5+
response = bedrock_runtime.invoke_model(+
modelId="global.anthropic.claude-opus-5-5",+
contentType="application/json",+
accept="application/json",+
body=json.dumps({+
"anthropic_version": "bedrock-2023-05-31",+
"max_tokens": 4096,+
"messages": [+
{+
"role": "user",+
"content": "An S3 bucket serves 40 TB/month egress. Estimate the monthly egress cost at $0.09/GB, and state one architecture change to cut it. Show the calculation, keep it under 120 words."+
}+
]+
})+
)+
+
result = json.loads(response["body"].read())+
# Opus 5.5 is a reasoning model: the response may include a thinking block+
# before the text block, so select the text block rather than a fixed index.+
print(next(b["text"] for b in result["content"] if b["type"] == "text"))</code></pre>+
</div>+
<p>You can also use the Amazon Bedrock Converse API for a unified multi-model experience:</p>+
<div class="hide-language">+
<pre><code class="language-python">import boto3+
+
# Create a Bedrock Runtime client+
bedrock_runtime = boto3.client(+
service_name="bedrock-runtime",+
region_name="us-east-1"+
)+
+
# Invoke Claude Opus 5.5+
response = bedrock_runtime.converse(+
modelId="global.anthropic.claude-opus-5-5",+
messages=[+
{+
"role": "user",+
"content": [+
{+
"text": "Can you explain the features of Amazon Bedrock?"+
}+
]+
}+
],+
inferenceConfig={+
"maxTokens": 4096+
}+
)+
+
if 'output' in response:+
blocks = response['output']['message']['content']+
print('\n'.join(b.get('text', '') for b in blocks if 'text' in b))</code></pre>+
</div>+
<p>You can also use the Anthropic Messages API through the <code>anthropic</code> SDK package for a streamlined experience:</p>+
<div class="hide-language">+
<pre><code class="language-python">from anthropic import Anthropic+
from aws_bedrock_token_generator import provide_token+
+
token = provide_token(region="us-east-1")+
+
client = Anthropic(+
base_url="https://bedrock-runtime.us-east-1.amazonaws.com/anthropic",+
api_key=token,+
)+
+
# Invoke Claude Opus 5.5+
response = client.messages.create(+
model="global.anthropic.claude-opus-5-5",+
max_tokens=1024,+
messages=[{"role": "user", "content": "Can you explain the features of Amazon Bedrock?"}],+
)+
print(response)</code></pre>+
</div>+
<p>You can explore the <a href="https://github.com/aws-samples/anthropic-on-aws/blob/main/notebooks/claude_fable_5_1_getting_started/claude-fable-5-1-getting-started.ipynb" target="_blank" rel="noopener">Getting Started notebook</a> for more examples. You can monitor usage, perfor…+
<h2 id="availability">Availability</h2>+
<p>Claude Opus 5.5 is available today on Amazon Bedrock through the US Geo CRIS (<code>us.</code>), EU Geo CRIS (<code>eu.</code>), AU Geo CRIS (<code>au.</code>), JP Geo CRIS (<code>jp.</code>) and Global CRIS (<code>global.</code>)…+
<p>See the <a href="https://docs.aws.amazon.com/bedrock/latest/userguide/model-cards-anthropic.html" target="_blank" rel="noopener">Amazon Bedrock documentation</a> for the full list of supported Regions. For pricing information, see <a href="https://aws.amazon.com/bedrock/pric…+
<p>Give Claude Opus 5.5 a try in the <a href="https://console.aws.amazon.com/bedrock" target="_blank" rel="noopener">Amazon Bedrock console</a>, in <a href="https://console.aws.amazon.com/claude-platform/" target="_blank" rel="noopener">Claude Platform on AWS</a>, or e…+
<hr style="width: 100%">+
<h2>About the authors</h2>+
<footer>+
<div class="blog-author-box" style="padding-top: 2.0em">+
<div class="blog-author-image" style="margin-right: 1.0em">+
<p><img loading="lazy" class="alignnone size-full" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/09/22/ML-22003-3.jpg" alt="Aamna Najmi" width="100" height="133"></p>+
</div>+
<h3 class="lb-h4">Aamna Najmi</h3>+
<p>Aamna is a Senior Specialist Solutions Architect for Generative AI focusing on Anthropic models and operationalizing and governing generative AI systems at scale on Amazon Bedrock. She helps ISVs solve their challenges, embrace innovation, and create new business opportunities with Amazo…+
</div>+
<div class="blog-author-box" style="padding-top: 2.0em">+
<div class="blog-author-image" style="margin-right: 1.0em">+
<p><img loading="lazy" class="alignnone size-full" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/09/22/ML-22003-4.jpg" alt="Dani Mitchell" width="100" height="133"></p>+
</div>+
<h3 class="lb-h4">Dani Mitchell</h3>+
<p>Dani is a Sr GenAI Specialist Solutions Architect at AWS and the SA lead for Amazon Bedrock Knowledge Bases. He helps enterprises across the world design and deploy generative AI solutions using Amazon Bedrock and Anthropic’s models and capabilities to build scalable, production-ready ap…+
</div>+
<div class="blog-author-box" style="padding-top: 2.0em">+
<div class="blog-author-image" style="margin-right: 1.0em">+
<p><img loading="lazy" class="alignnone size-full" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/09/22/ML-22003-5.jpg" alt="Sofian Hamiti" width="100" height="133"></p>+
</div>+
<h3 class="lb-h4">Sofian Hamiti</h3>+
<p>Sofian is a technology leader with over 12 years of experience building AI solutions, and leading high-performing teams to maximize customer outcomes. He is passionate about empowering diverse talents to drive global impact and achieve their career aspirations.</p>+
</div>+
<div class="blog-author-box" style="padding-top: 2.0em">+
<div class="blog-author-image" style="margin-right: 1.0em">+
<p><img loading="lazy" class="alignnone size-full" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/09/22/ML-22003-6.jpg" alt="Eugenio Soltero" width="100" height="133"></p>+
</div>+
<h3 class="lb-h4">Eugenio Soltero</h3>+
<p>Eugenio is a Sr.&nbsp;Product Marketing Manager for Amazon Bedrock at AWS. With several years of experience in generative AI, he helps customers navigate the evolving landscape of foundation models and generative AI to adopt solutions that deliver measurable value.</p>+
</div>+
</footer>+
</div></content:encoded>+
+
+
+
</item>+
<item>+
<title>Evaluate skill-equipped agents with Strands Evals and Amazon Bedrock AgentCore</title>+
<link>https://aws.amazon.com/blogs/machine-learning/evaluate-skill-equipped-agents-with-strands-evals-and-amazon-bedrock-agentcore/</link>+
+
<dc:creator><![CDATA[Sangmin Woo]]></dc:creator>+
<pubDate>Tue, 22 Sep 2026 17:18:13 +0000</pubDate>+
<category><![CDATA[Advanced (300)]]></category>+
<category><![CDATA[Amazon Bedrock AgentCore]]></category>+
<category><![CDATA[Announcements]]></category>+
<guid isPermaLink="false">aa00d9c221ad9292cfbcc5630acd0f904e2648ad</guid>+
+
<description>Skills let you encode domain-specific procedures as reusable, portable instructions for agents, but a fluent answer doesn't prove the agent picked the right skill or followed it. Learn how to measure skill selection and instruction following with Strands Evals and Amazon Bedrock Ag…+
<content:encoded><p>General-purpose agents handle a broad range of tasks, but you still need them to follow the procedures that run your business: compliance checks, document-processing workflows, escalation policies, engineering conventions. Encoding all of that in one system prompt…+
<p>A skill packages one or more tools with the context an agent needs to use them correctly:</p>+
<ul>+
<li><strong>Instructions</strong>: Domain-specific guidance and constraints injected into the agent’s context.</li>+
<li><strong>Tool bindings</strong>: The APIs, Model Context Protocol (MCP) servers, or local commands the skill depends on.</li>+
<li><strong>Knowledge</strong>: Reference material and worked examples.</li>+
<li><strong>Workflow</strong>: The multi-step procedure or decision logic the skill follows.</li>+
<li><strong>Guardrails</strong>: Format requirements, scope limits, and validation rules.</li>+
</ul>+
<p>This modular approach helps teams specialize agents faster, reuse proven procedures across agents and workflows, keep behavior consistent, and update domain-specific guidance without fine-tuning the underlying model or rewriting the agent’s core logic.</p>+
<p>This skill composability in agents introduces two failure modes that general output-quality metrics can miss: the agent invokes a skill that is not appropriate for the task and the agent invokes the right skill but skips or only partially follows its instructions. Both failures can produce …+
<p>To make these failures measurable, <a href="https://strandsagents.com/docs/user-guide/evals-sdk/quickstart/" target="_blank" rel="noopener">Strands Evals SDK</a> and <a href="https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/evaluations.html" target="_blank" rel="…+
<ul>+
<li>Skill Selection Accuracy determines whether each invoked skill was an appropriate choice for the task. It returns a binary result for each invoked skill.</li>+
<li>Skill Instruction Following determines how fully the agent followed an invoked skill’s instructions. It returns a five-level rating grounded in evidence for each prescribed step.</li>+
<li>Additionally on Strands Evals, Skill Invoked is a deterministic check of whether a named skill has been loaded successfully.</li>+
</ul>+
<p>In this post, you will learn how to evaluate skill selection and instruction following from a recorded trajectory in Strands Evals, add deterministic routing checks to a test suite, evaluate skill behavior from OpenTelemetry traces with AgentCore Evaluations, and interpret per-skill results…+
<h2 id="understand-what-each-evaluator-measures">Understand what each evaluator measures</h2>+
<p>An agent receives a task and a catalog of skills, chooses a skill, loads it, and acts. The run is recorded as a trajectory in Strands Evals or an OpenTelemetry trace in your observability layer. This record can now be used for all three skill evaluators. <a href="https://docs.aws.amazon.…+
<div style="width: 810px" class="wp-caption alignnone">+
<a href="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/09/16/ML-21794-1.png" target="_blank" rel="noopener"><img src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/09/16/ML-21794-1.png" alt="Diagram of an agent sel…+
<p class="wp-caption-text">Figure 1: Skill Selection Accuracy checks whether the agent chose an appropriate skill for the task</p>+
</div>+
<p><a href="https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/skill-evaluators.html#skill-evaluators-sif" target="_blank" rel="noopener">Skill Instruction Following</a> asks how fully the agent followed that skill’s prescribed steps. You can find the prompt template and…+
<div style="width: 810px" class="wp-caption alignnone">+
<a href="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/09/16/ML-21794-2.png" target="_blank" rel="noopener"><img src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/09/16/ML-21794-2.png" alt="Diagram of Skill Instru…+
<p class="wp-caption-text">Figure 2: Skill Instruction Following measures how completely the agent followed the skill’s steps</p>+
</div>+
<p><code>SkillInvoked</code> is deterministic. It calls no model and is specific to Strands Evals.</p>+
<p>An overview of these skill evaluators is demonstrated diagrammatically in the following figure.</p>+
<div style="width: 810px" class="wp-caption alignnone">+
<a href="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/09/16/ML-21794-3.png" target="_blank" rel="noopener"><img src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/09/16/ML-21794-3.png" alt="Overview diagram of the…+
<p class="wp-caption-text">Figure 3: Overview of the three skill evaluators</p>+
</div>+
<p>Consider an <a href="https://github.com/awslabs/agentcore-samples/tree/main/01-features/06-observe-evaluate-optimize-your-agent/02-evaluate/skills-evaluation" target="_blank" rel="noopener">HR assistant agent</a> with skills for paid time off (PTO) planning and discussing employ…+
<p>Now suppose the agent correctly invokes the PTO-planning skill for a related request. The skill instructs the agent to identify the <code>employee_id</code>, check the PTO balance, check the rollover rules against the latest HR policy, and then submit a PTO request if the condit…+
<p>The failures require different fixes. An inappropriate selection often points to overlapping or ambiguous skill descriptions. Incomplete instruction following might call for clearer steps, a different skill structure, or a more capable agent model.</p>+
<table border="1px" width="100%" cellpadding="10px"> <tbody> <tr>-
<td><strong>Evaluation</strong></td>-
<td><strong>Grok 4.6 High</strong></td>-
</tr>-
<tr>-
<td>AA Intelligence Index</td>-
<td>61</td>+
<td><strong>Evaluator</strong></td>+
<td><strong>Availability</strong></td>+
<td><strong>Score</strong></td>+
<td><strong>Question answered</strong></td> </tr> <tr>-
<td>GDPVal-AA v2</td>-
<td>1753</td>+
<td><strong>Skill Selection Accuracy</strong></td>+
<td>Strands Evals and AgentCore Evaluations</td>+
<td>Binary, per invoked skill</td>+
<td>Was invoking this skill appropriate for the task?</td> </tr> <tr>-
<td>CursorBench v3.2</td>-
<td>69.9%</td>+
<td><strong>Skill Instruction Following</strong></td>+
<td>Strands Evals and AgentCore Evaluations</td>+
<td>Five levels, per invoked skill</td>+
<td>How fully did the agent follow this skill’s prescribed steps?</td> </tr> <tr>-
<td>DeepSWE v1.1</td>-
<td>65.9%</td>+
<td><strong>Skill Invoked</strong></td>+
<td>Strands Evals</td>+
<td>Binary, deterministic</td>+
<td>Was this named skill successfully loaded?</td> </tr>+
</tbody>+
</table>+
<p>Because judge-based evaluators return per-invoked-skill results, multi-skill runs remain diagnosable: you can identify which selection or instruction-following result lowered the aggregate score. If no skill is invoked, the judge-based evaluators don’t produce a score. Pair them with <co…+
<h2 id="prerequisites">Prerequisites</h2>+
<ul>+
<li>Python 3.10 or later.</li>+
<li>An AWS account with Amazon Bedrock access, and credentials with <code>InvokeModel</code> permission for the judge model.</li>+
</ul>+
<p><strong>To follow the Strands Evals section, install the SDKs:</strong></p>+
<div class="hide-language">+
<pre><code class="language-bash">pip install strands-agents-evals strands-agents</code></pre>+
</div>+
<p>You also need a recorded agent run. The skill evaluators accept either a Strands Evals Session or a raw message list as the trajectory. At launch, skill extraction recognizes signals from the Strands AgentSkills plugin, Claude Code, Claude Agent SDK, OpenAI Agents SDK, Codex, Gemini CLI, Op…+
<p><strong>To follow the AgentCore Evaluations section, you need:</strong></p>+
<ul>+
<li>An agent hosted on Amazon Bedrock AgentCore runtime or elsewhere. We will use the example of the <a href="https://github.com/awslabs/agentcore-samples/blob/main/01-features/06-observe-evaluate-optimize-your-agent/02-evaluate/utils/hr_assistant_agent.py" target="_blank" rel="noopener"&g…+
<li>Observability enabled for that agent, so it delivers telemetry to Amazon CloudWatch.</li>+
<li>Transaction Search enabled in CloudWatch.</li>+
</ul>+
<p>The examples in this post use the AgentCore CLI:</p>+
<div class="hide-language">+
<pre><code class="language-bash">npm install -g @aws/agentcore</code></pre>+
</div>+
<h2 id="evaluate-a-recorded-trajectory-with-strands-evals">Evaluate a recorded trajectory with Strands Evals</h2>+
<p>Strands Evals is useful when you control the test cases and can rerun the agent during development or continuous integration. Check out the complete Strands evals code sample created for the HR assistant agent in the <a href="https://github.com/awslabs/agentcore-samples/blob/main/01-feat…+
<h2 id="define-the-case-and-evaluators">1. Define the case and evaluators</h2>+
<div class="hide-language">+
<pre><code class="language-python">from strands_evals import Case, Experiment+
from strands_evals.evaluators import (+
SkillInstructionFollowingEvaluator,+
SkillInvoked,+
SkillSelectionAccuracyEvaluator,+
)+
+
case = Case(+
name="q3-revenue-tables",+
input="Summarize the revenue tables in q3-report.pdf",+
)+
+
evaluators = [+
SkillSelectionAccuracyEvaluator(),+
SkillInstructionFollowingEvaluator(),+
SkillInvoked(skill_name="pdf-table-extraction"),+
]</code></pre>Diff display stops at 400 lines. The line counts above are from the whole diff. 49 lines shown here cut at 300 characters. The raw artifact at this commit is linked above.