Jump to content

Draft:Amazon Nova

From Wikipedia, the free encyclopedia

Amazon Nova is a family of large language models developed by Amazon Web Services.[1]

Foundation models

[edit]

Amazon Nova large language models were released in four varying sizes: Micro, Lite, Pro, and Premier. Micro only has text-input, but the other three models can also take images and videos as input.[1]

Nova Canvas

[edit]

Amazon Nova Canvas is an image generation model.[2]

Nova Reel

[edit]

Amazon Nova Reel is a video generation model. Nova Reel 1.1, a newer version, was announced in April 2025.[3]

Nova Act

[edit]

Nova Act, based on the Amazon Nova foundation models and receiving additional reinforcement learning training, is an AI "agent".[4]

References

[edit]
  1. ^ a b Wiggers, Kyle (3 December 2024). "Amazon announces Nova, a new family of multimodal AI models". TechCrunch. Retrieved 23 April 2025.
  2. ^ Peters, Jay (4 December 2024). "Amazon announces its own set of Nova AI models". The Verge. Retrieved 25 April 2025.
  3. ^ Wiggers, Kyle (7 April 2025). "Amazon says its AI video model can now generate minutes-long clips". TechCruch. Retrieved 25 April 2025.
  4. ^ Knight, Will (31 March 2025). "Amazon's AGI Lab Reveals Its First Work: Advanced AI Agents". Wired. Retrieved 25 April 2025.