What AI model does it use?

What AI model does Alexa+ use?

Amazon has been consistent about this for a year and a half, and consistently vague. Alexa+ runs on two families of large language models, Amazon's own and Anthropic's. Amazon has never gone further than naming the families, in every Alexa+ post of its own that I read.

A router between Nova and Claude

The clearest description is in Amazon's engineering post from the February 26, 2025 launch. To balance speed against accuracy, Amazon wrote, "we built a sophisticated routing system using state-of-the-art models from Amazon Bedrock", "including Amazon Nova and Anthropic Claude", "instantly matching each customer request with the best model for the task at hand." (Bedrock is the AWS service that hosts other companies' AI models as well as Amazon's; Nova is Amazon's own model line.)

So every time you talk to Alexa+, a piece of Amazon's software decides which model answers. You don't choose. I couldn't find any setting or post that describes a way to.

Anthropic confirmed it the same day: "Today, we're announcing that Claude models are helping power Alexa+." No version there either.

Still both, in September 2026

A lot has happened to both companies' models since early 2025, so I wanted to know if the wording had moved. It hasn't. Amazon's Alexa+ availability post, first published in February 2026 and updated on September 1 (a note at the top says the update was about tier details), still describes "large language models", "from both Amazon Nova and Anthropic." That's the newest first-party wording I found.

Pricing doesn't come with a model difference, at least none Amazon mentions, and I'm not sure whether the free text chat reaches the same models as the voice assistant. Alexa+ is included with Prime in the US and costs $19.99 a month without it. There's also a free, text-only Alexa+ chat at Alexa.com and in the Alexa app for anyone (with use limits).

Why there's no version number

What follows is my own reading, since Amazon never explains it. A router makes a version number less meaningful: if software picks between several models per request, and Amazon can swap models behind it as new ones arrive, there's no single "Alexa+ model" to announce. Cursor, GitHub Copilot and Perplexity have all moved toward the same shape with their Auto or Best settings.

What's still unknown

Which Nova and which Claude models run it, which requests go to which, and whether the balance has shifted since 2025. News reports have described a split between the two, but they lean on unnamed sources, so I've kept them off this page.