News
ReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior ReleaseReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior Release
EveryLLM
imgAmazon Nova

Amazon Nova

Amazon's family of foundation models on AWS Bedrock

Amazon Nova is a family of foundation models built by Amazon AGI and offered through Amazon Web Services. The models were introduced in 2024 and a technical report and model card was published in 2025, covering text and multimodal variants designed for different cost and capability trade-offs.

Share on WhatsApp

  • Family of foundation models built by Amazon AGI
  • Introduced in 2024 and documented in a 2025 technical report
  • Includes text and multimodal variants
  • Served through Amazon Bedrock as a managed API
  • Weights are proprietary, not downloadable

What it is

Amazon Nova is a group of foundation models created by Amazon AGI and made available through Amazon Web Services. The company introduced them in 2024 and describes them as offering strong capability at competitive prices. The family includes models aimed at different jobs, from lightweight text work to multimodal tasks that accept images alongside text. A technical report and model card published in 2025 documents the family in detail. Amazon positions Nova as part of its Bedrock platform, where customers can call the models through an API rather than running them themselves.

How it came about

Amazon announced the Nova family in 2024 at its re:Invent event, presenting them as its own foundation models for AWS customers. In 2025 the Amazon AGI team released a technical report and model card on arXiv, listing a very large author team and describing the models' design, training and evaluation. The report covers the family as a whole rather than a single model.

Inside the model

The technical report describes Nova as a family of models built by Amazon AGI, with variants sized for different workloads and some able to take images as well as text. The published documentation covers training approach, evaluation and model cards for the family. Detailed figures on layers, parameter counts, context windows and training compute are given in the report itself.

What it can do

Amazon says the Nova models deliver strong performance relative to their price, and the family spans text-only and multimodal tasks. The technical report includes benchmark results and model cards for the different variants. Specific scores are reported in the report rather than summarised here.

How to use it

Nova models are offered through Amazon Bedrock as a managed service, so customers use them via API inside AWS. They are proprietary models; Amazon does not release the weights for download.

Safety and testing

The technical report and model card published by Amazon AGI document evaluation of the models. Further detail on safety work sits in that documentation.

Reception and debate

Amazon presented Nova as a price-performance play for AWS customers, and the family was covered in the company's own announcement material.

Why it matters

Nova gives AWS customers a first-party foundation model option alongside the third-party models on Bedrock, and the 2025 technical report adds to the public documentation of large commercial model families.

Questions people ask about Amazon Nova

What is Amazon Nova?

Nova is Amazon's own family of foundation models, built by Amazon AGI and offered to customers through Amazon Web Services. It includes several variants, some text-only and some that also accept images, aimed at different capability and cost points.

Who makes it?

Amazon AGI, Amazon's artificial general intelligence team, develops the models. They are delivered through Amazon Web Services, and a technical report and model card from the team was published in 2025.

How do I use it?

Through Amazon Bedrock, AWS's managed platform for foundation models. Customers call the models by API inside AWS rather than downloading weights, since the models are proprietary.

Is it open source?

No. Nova is a proprietary family served through Amazon Bedrock. Amazon has published a technical report and model card describing the models, but the weights are not released for download.

Good reading and sources

    Where this page comes from

    There is no English Wikipedia article on Amazon Nova yet, so the facts on this page come from the sources listed above: aws.amazon.com/blogs/aws/introducing-amazon-nova-frontier-intelligence, arxiv.org/abs/2506.12103. Some are the subject's own pages, so its claims about itself are reported as its claims. EveryLLM wrote the text in its own words, with AI assistance, and compares it with the sources to make sure no passage is copied.

    It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

    The weekly digest

    Every Monday: the model of the day, new pages on EveryLLM, and the conferences and anniversaries of the week ahead.