Skip to main content

Amazon unveils its new family of Nova foundational models

pasta city
AWS

Amazon CEO Andy Jassy took to the stage at the company’s re:Invent conference on Tuesday to show off six new text, image, and video generation models that it calls Amazon Nova.

This new family of multimodal generative AIs includes Nova Micro, a text-only model built for low-cost, low-latency responses; Nova Lite, a low-cost multimodal model for processing image, video, and text inputs; and Nova Pro, its general purpose multimodal model that combines “accuracy, speed, and cost for a wide range of tasks,” per the company’s announcement post. Nova Premier is Amazon’s “most capable … multimodal models for complex reasoning tasks,” while Nova Canvas is a dedicated text-to-image engine and Nova Reel is purpose-built to generate video.

Recommended Videos

The text-based models have been optimized on 15 different languages. Micro offers a 128,000-token context window while both Lite and Pro can handle up to 300,000 tokens (around 225,000 words or 30 minutes of video). The company plans to expand the context windows of its larger models up to 2 million tokens by early next year. 

Canvas enables users to generate and edit images using natural language prompts. Reels, which will compete with the likes of Gen-3 Alpha, Kling, and Dall-E 3, can generate clips up to six seconds in length from both text prompts and reference images. The video generator also offers camera motion control including pans and zooms.

Pasta City, created with Amazon Nova Reel by Amazon Ads

“We’ve continued to work on our own frontier models,” Jassy told the assembled crowd, “and those frontier models have made a tremendous amount of progress over the last four to five months. And we figured, if we were finding value out of them, you would probably find value out of them.”

Jassy also says that these models are both among the least expensive to operate and fastest in their class, though the company has yet to post benchmark data supporting those claims. “We’ve optimized these models to work with proprietary systems and APIs, so that you can do multiple orchestrated automatic steps — agent behavior — much more easily with these models,” he said. “So I think these are very compelling.”

The Micro, Lite, and Pro models (as well as Canvas and Reels) are all currently available to AWS customers. Premiere is set to arrive in Q1 2025.

Andrew Tarantola
Andrew Tarantola is a journalist with more than a decade reporting on emerging technologies ranging from robotics and machine…
ChatGPT unveils Sora with up to 20-second AI video generation
An AI generated image of a woman who walks the streets of Tokyo.

OpenAI has been promising to release its next-gen video generator model, Sora, since February. On Monday, the company finally dropped a working version of it as part of its "12 Days of OpenAI" event.

"This is a critical part of our AGI roadmap," OpenAI CEO Sam Altman said during the company's live stream.

Read more
ChatGPT’s new Pro subscription will cost you $200 per month
glasses and chatgpt

Sam Altman and team kicked off the company's "12 Days of OpenAI" event Thursday with a live stream to debut the fully functional version of its 01 reasoning model, as well as a new subscription tier called ChatGPT Pro. But to gain unlimited access to these new features and capabilities, you're going to need to shell out an exorbitant $200 per month.

The 01 model, originally codenamed Project Strawberry, was first released in September as a preview, alongside a lighter-weight o1-mini model, to ChatGPT-Plus subscribers. o1, as a reasoning model, differs from standard LLMs in that it is capable of fact-checking itself before returning its generated response to the user. This helps such models reduce their propensity to hallucinate answers but comes at the cost of a longer inference period and slower response.

Read more
Windows 11 Recall finally arrives, but with one new problem
Recall screenshot.

The Windows 11 Recall feature has been troublesome since its announcement. Now that the feature is available in a testing capacity, it is still causing users some issues.

Tech reviewers testing the feature have observed that Recall will now fail to save the snapshots that allow the function to work. CNBC noted that it may take “several minutes” for a snapshot to save, which may leave delays in the AI processing. Tom Warren of The Verge, noted on Bluesky that snapshots were not saving at all in his experience.

Read more