AWS AI

"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.

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.

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

Near-Astra intelligence, built for everyday work. GPT-6.1 Sol from @OpenAI is now generally available on Amazon Bedrock.

Build and scale capable agents in production with the performance, security, and reliability of Amazon Bedrock.

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.