eWEEK content material and product suggestions are editorially impartial. We might generate profits whenever you click on on hyperlinks to our companions. Learn More.
At its re:Invent 2024 convention, Amazon’s AWS unveiled Nova, a cutting-edge household of AI fashions designed to revolutionize how companies leverage AI. This announcement marks a major leap in Amazon’s AI improvement, spotlighting six generative AI fashions designed to simplify textual content, picture, and video content material era. These AI instruments are accessible to enterprise purchasers through Amazon Bedrock.
“Inside Amazon, we have about 1,000 Gen AI applications in motion, and we’ve had a bird’s-eye view of what application builders are still grappling with,” Amazon SVP of Artificial General Intelligence Rohit Prasad mentioned in an announcement. “Our new Amazon Nova models are intended to help with these challenges for internal and external builders, and provide compelling intelligence and content generation while also delivering meaningful progress on latency, cost-effectiveness, customization, information grounding, and agentic capabilities.”
Meet the Nova Models
The Nova sequence contains 5 presently obtainable fashions and a sixth, Nova Premier, slated for launch in Q1 2025. The AI fashions are constructed to deal with numerous enterprise wants, from textual content to superior multimedia content material era.
Text-only fashions embrace Micro, Lite, and Pro:
- Nova Micro: Ideal for fast text-to-text duties, Micro is the smallest within the lineup however boasts the bottom latency. Its context window of 128Ok tokens ensures effectivity for duties comparable to textual content summarization, translation, content material classification, interactive chat and brainstorming, and easy mathematical reasoning and coding.
- Nova Lite: The mannequin accepts pictures, movies, and textual content as inputs however solely generates textual content. It processes inputs as much as 300Ok tokens in size and may analyze a number of pictures or as much as 30 minutes of video in a single request.
- Nova Pro: The most strong among the many trio, Pro delivers superior accuracy and multitasking capabilities. It additionally encompasses a 300Ok-token capability, making it appropriate for advanced workflows.
Creative content material fashions embrace Canvas and Reel:
- Nova Canvas: Specializing in picture creation, Canvas can generate visuals primarily based on textual content or picture prompts. It’s designed for industries like promoting and leisure, with instruments for coloration changes and background edits.
- Nova Reel: A video era powerhouse, Reel produces brief clips from textual content and picture prompts. Users can manipulate digicam actions utilizing pure language instructions, making it a super device for entrepreneurs and content material creators.
- Nova Premier: The most superior mannequin, constructed for advanced reasoning and fine-tuning customized AI options. Amazon Nova Premier remains to be in coaching. It’s set to launch in early 2025.
What Makes Nova Different?
One standout characteristic is Nova’s deal with value effectivity. Amazon Nova Micro, Amazon Nova Lite, and Amazon Nova Pro are a minimum of 75 p.c cheaper than the best-performing fashions of their respective intelligence courses in Amazon Bedrock, Amazon mentioned. They additionally excel at customization, permitting companies to fine-tune fashions utilizing their proprietary knowledge for improved accuracy and relevance.
In 2025, some Nova fashions’ context home windows will broaden to help over two million tokens, the equal of greater than 1.5 million phrases, marking one other leap ahead. Additionally, AWS plans to launch a speech-to-speech mannequin in early 2025, promising natural-sounding transformations by decoding tone and cadence. An bold any-to-any mannequin, combining inputs and outputs throughout textual content, speech, pictures, and video, is…
