Model components / Fine-tuned / chat models

Amazon Nova Pro

Amazon's multimodal model line served through AWS Bedrock. The current generation is Nova 2, launched at re:Invent in December 2025, spanning a Pro tier, a Lite speed-and-cost tier, an Omni tier taking text, image, video and audio in, and a Sonic speech-to-speech tier. Nova 2 Pro emphasizes agentic capability over the original line.

Scores Nova Pro v1, a multimodal understanding model with a 300K context reached only through Bedrock and still in service under Nova 2. AWS publishes no standardized benchmark scores, so capability abstains. Verified 2026-08-13 via the AWS Nova user guide and models page.

Openness

1 high confidence

1.0

weights: closed
data: closed
code: closed
license: Proprietary(API-only via Amazon Bedrock)

Served only through Amazon Bedrock with no weights released, so fully closed. Nova Pro is the release the tier is read on; Micro, Lite and Pro are tiers within one family Amazon markets as Nova, which is why they are recorded together.

https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html recorded 2026-08-13
"To start building with Amazon Nova, you must access the models through an API using Amazon Bedrock"; Nova Pro described as a multimodal understanding model; no weights, corpus or training code

Adoption

4 low confidence

4.0

Distributed broadly through Amazon Bedrock across many AWS regions (with fine-tuning, provisioned throughput, Bedrock Agents/KB support) and positioned as the price-performance default of the Nova family. No standalone per-model user/usage count published; level 4 reflects Bedrock's enterprise distribution surface rather than a verified figure.

https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html recorded 2026-08-13
"Amazon Nova Pro ... Highly capable multimodal model with best combination of accuracy, speed, and cost for a wide range of tasks", listed across AWS regions on Bedrock; no user or usage count published

Capability

3 low confidence

3.0

Amazon's live product page publishes no standardized benchmark scores at all - the word "benchmark" does not appear on it - so the positioning rests on price-performance claims and case studies. That is a searched absence rather than an unexamined one. Mid-tier against the 2026 frontier, scored 3 directionally.

https://aws.amazon.com/nova/models/ recorded 2026-08-13
price-performance and customization positioning; the word "benchmark" does not appear on the page, so no standardized score is published for the axis to read

Verified 2026-08-13