AWS subscribers now have entry to generative AI fashions that compete with GPT-4o. On December 3, on the AWS re:Invent occasion in Las Vegas and on-line, AWS introduced six new mannequin sizes for various use instances within the new Amazon Nova household.
“Within Amazon, we’ve about 1,000 generative AI purposes on the go, and we have had a hen’s-eye view of what software builders are nonetheless grappling with,” stated Rohit Prasad, SVP of Amazon Artificial General Intelligence , In the press release.
“Our new Amazon Nova fashions are designed to assist inside and exterior builders overcome these challenges and supply intelligence and compelling content material technology, whereas delivering important advances in latency, cost-effectiveness, personalization, retrieval Augmented Generation (RAG) and agent capabilities. “
What is Amazon Nova?
Amazon Nova is a line of baseline fashions for generative synthetic intelligence obtainable on AWS’ Amazon Bedrock AI internet hosting service. Today, organizations can experiment with three sizing choices:
- Amazon NovaMicro it’s a text-only mannequin with a quick response time of 210 output tokens per second. Amazon claims to outperform Meta’s Llama 3.1 8B and Google’s Gemini 1.5 Flash-8B. Nova Micro is meant for purposes that require speedy responses at a comparatively low value.
- Amazon Nova Lite is one other small mannequin from the Nova household. Unlike Micro, it could actually analyze picture, video or textual content enter. Comparable to OpenAI’s GPT-4o mini, Nova Lite is meant for speedy summarization and interpretation of graphs or video shows. Because it could actually perceive photos on pc screens and execute operate calls, Amazon Nova Lite is suitable for some quasi-autonomous chained habits used for “AI agent” duties.
- Amazon Nova Pro it’s the mid-range mannequin. Amazon stated it really works quicker, extra precisely and prices lower than OpenAI’s GPT-4o or Google’s Gemini 1.5 Pro. Nova Pro can interpret textual content or photos and helps agent workflows.
Once clients have a Nova mannequin, they will refine it primarily based on their proprietary knowledge.
In addition to dimension choices, organizations can even select between a picture technology template (Amazon Nova Canvas) and a video template (Amazon Nova Reels). Both are meant to create “studio high quality” content material.
- New canvas Create photos primarily based on textual content or picture ideas. Amazon notes that it consists of safety features like watermarking and content material guardrails.
- Nova Reels creates six-second movies, with Amazon planning to increase the attainable video size to 2 minutes within the “coming months.”
SEE: AI regulation is underway in Australia, with a committee calling for giant fashions from OpenAI, Meta and Google to be deemed “excessive danger.”
What’s the subsequent step?
The fourth mannequin within the Nova line, Nova Premier, will not be obtainable till the primary quarter of 2025. Amazon expects Nova Premier to supply multimodal interpretation (video, picture, or text-to-text) and an in depth knowledge library that organizations they will use to coach different fashions.
Additionally, Amazon plans so as to add a mannequin that may reply naturally to spoken dialog. They are additionally engaged on a multimodal to multimodal mannequin for decoding and producing textual content, photos, video or audio.
While it is nonetheless too early to see how Nova will compete with rivals like OpenAI, Google and Meta, Amazon has landed a significant companion in SAP, which gives fashions on its AI Core platform.
TechRepublic covers AWS re:Invent remotely.
