"There is no one model to rule them all."
The industry says AI models are becoming commoditized. Matt Wood says that's just not playing out.
Every model has its own reasoning patterns, biases, & sweet spots. In an agent workflow running hundreds of interactions at machine speed, even a 1% difference compounds into a completely different direction. The case for model choice is stronger than most people think.
1
570
Claude Haiku 5.5 is now available on AWS.
The fastest, most efficient model in the Claude 5.5 family, and @AnthropicAI's most capable Haiku yet, built for subagents and high-volume, cost-sensitive work. It also costs around 75% less than Claude Haiku 4.5 for most tasks.
Available on Amazon Bedrock and Claude Platform on AWS. go.aws/4ydOjJl
306
Two new AI League challenges are live. Pick yours and compete for a shot at re:Invent.
Customize a Foundational Model: fine-tune a model on Amazon SageMaker AI to answer re:Invent questions like a seasoned veteran.
Agentic Football Cup: fine-tune a model, then deploy a 5-agent squad on Amazon Bedrock AgentCore to compete live.
Both challenges post to a live regional leaderboard. Top scorers advance to a live regional finale, and each regional winner heads to the Grand Finale at AWS re:Invent 2026.
go.aws/3Uf4Oqp
Learn more. go.aws/3VXML8x
371
AI models catch nearly every vulnerability. They also flag safe code as dangerous up to 99% of the time.
We've released the Deception Benchmark: 14,822 code samples designed to measure whether AI models can tell real vulnerabilities from safe code that merely looks suspicious.
Explore the benchmark. go.aws/3Vgk2vs
1
761
We looked at what successful leaders do differently. go.aws/4rvqxGF
1
1,241
Learn more. go.aws/4iIE8bB
1
643
Introducing Amazon CloudWatch Omni, the next evolution of Amazon CloudWatch, a re-imagined observability experience for application and AI workloads that is AI-powered, built on open standards, and delivered where you work.
Your applications, your agents, your infrastructure. One place.
Learn more: go.aws/3TLqGth
2
889
Replying to @claudeai
@claudeai Opus 5.5 is now on AWS, @AnthropicAI's most capable Opus yet at a lower cost per token.
The enterprise workhorse for long-running coding and knowledge work.
Available on Amazon Bedrock and Claude Platform on AWS.
Learn more: go.aws/4rqjQ8M
Learn More. go.aws/4yaLQAb
2
623
Explore and deploy the newest AI agents developed through PAF co-innovation in AWS Marketplace.
Learn more. go.aws/4iHyXJ1
2
1
455
Read the full August recap. August's AI builder updates on AWS are live. go.aws/4iSIwof
1
300
The AWS AI League opening challenge is now open.
Build and deploy an AI agent on Amazon Bedrock AgentCore that navigates the challenge, answers questions, and reaches the objective, fast.
Live leaderboard.
Top scorers face off in a regional finale. The regional winner heads to the re:Invent Grand Finale.
Claude Fable 5.1, Anthropic’s new frontier model, is now available on Amazon Bedrock and Claude Platform on AWS.
Fable 5.1 is built for production work across your code, your documents, and your research. A step change in reasoning on the hardest problems. Powers agents across the software lifecycle in hours-long sessions, with honest reporting when work stalls. Knowledge work from question to finished deliverable. Research support from idea to validation
Read the blog. go.aws/4gLc7xe
Learn more. go.aws/3UnlN9L
1
223
Announcing: Ray on Amazon SageMaker HyperPod
Run distributed AI at production scale with managed development environments, built-in observability, resilient training, and accelerated serving.
Fully compatible with open-source Ray, running on Amazon SageMaker HyperPod clusters orchestrated with Amazon EKS.
Your code. Our infrastructure. Zero ops tax.
go.aws/4cTZTBd
OpenAI GPT-5.4 is now on Amazon Bedrock in AWS GovCloud (US). FedRAMP High & DoD IL4/5 authorized.
Federal builders: access frontier AI within your compliance boundary.
Start building. go.aws/4hE9kbn



















