AWS
Amazon Nova
the Bedrock native
Key facts
- AmazonAWS first-party
- Maker
- Bedrocknative option
- Platform
- Several sizescost vs capability
- Range
- Everyday workloadspriced low
- Positioning
- AWS customerssame cloud boundary
- Best for
The Bedrock native. Amazon's own foundation family on Bedrock.
What Amazon Nova is
Amazon Nova is Amazon’s own family of foundation models, and its defining feature is where it lives: natively inside Amazon Bedrock, the managed AWS service through which customers reach a range of AI models. Bedrock hosts systems from several providers, yet Amazon Nova is the house option, built by Amazon and offered as the first-party choice for the many businesses that already run on Amazon Web Services.
What Bedrock is
Bedrock is worth explaining, because it frames everything about the model. Rather than train their own systems, most companies want to call an existing one through a familiar, secure interface, with billing, access controls and data handling that fit the rest of their cloud setup. Bedrock provides exactly that: a single managed service on AWS for reaching foundation models without running any infrastructure. By placing its own family at the centre of Bedrock, Amazon gives cloud customers an in-house option that needs no separate contract or integration.
The native advantage
The advantage of being native to Bedrock is mostly practical. A team already using AWS can adopt Amazon Nova with the permissions, monitoring and billing they know, keeping their data inside the same cloud boundary. For large organisations that value that continuity, an in-house model integrated this tightly becomes an easy default, even before questions of raw capability come into play. Amazon has priced and positioned the family to be the sensible starting point for Bedrock users rather than a bid for the outright performance crown.
There is a cost dimension too. Because Amazon controls the whole stack, from the chips in its data centres to the Bedrock service on top, it can offer its own models at prices that are hard for an outside provider to match, and it has leaned on that advantage in positioning the family for everyday workloads. For a business watching its cloud bill, a capable model that is cheap to run and already integrated is a strong proposition.
A family of sizes
Amazon Nova is a family rather than one model, offered in several sizes and configurations so that customers can trade capability against cost. Lighter versions suit high-volume, routine work where speed and price dominate, while more capable versions handle harder tasks. This tiering mirrors how the rest of the field is structured, and it lets a Bedrock customer keep everything within the same first-party family as their needs change.
The Bedrock setting also shapes how the family is judged. A buyer is rarely choosing the model in isolation; they are choosing it against the other systems available on the same platform, which makes convenience, price and integration count for as much as headline capability. That is the ground Amazon has chosen to compete on, and it plays to the strengths of a company that sells infrastructure first and models second.
Where it sits
Because Amazon keeps developing the line, anyone deploying it should confirm the current generation rather than assume a fixed set of capabilities, as the family has been extended since its introduction. The strategic point holds across versions: this is Amazon’s move to keep model usage, and the revenue that comes with it, inside its own cloud rather than flowing to an outside provider hosted on Bedrock.
Where Amazon Nova sits in the wider field is as the cloud giant’s answer to a simple commercial question. If customers are going to run AI on AWS anyway, Amazon would rather they run Amazon’s own model. That makes the family a quiet default for enterprise buyers already committed to Bedrock rather than a contender for public mindshare. For how it compares with the independent model families it shares that platform with, see our large language models hub and the wider AI section.
More in Large Language Models
All LLMs →- OpenAIGPT-5.6current flagship
- AnthropicClaude Fable 5current flagship, top of most July 2026 rankings
- AnthropicClaude Mythos 5restricted twin of Fable 5
- Google DeepMindGemini 3.5 familycurrent Google generation
- xAIGrok 4.5current xAI flagship, coding pivot
- Moonshot AIKimi K3the week's big story, largest open-weight model ever