Proper on the heels of saying Nova Forge, a service to coach customized Nova AI fashions, Amazon Net Companies (AWS) introduced extra instruments for enterprise clients to create their very own frontier fashions.
AWS introduced new capabilities in Amazon Bedrock and Amazon SageMaker AI at its AWS re:Invent convention on Wednesday. These new capabilities are designed to make constructing and fine-tuning customized massive language fashions (LLMs) simpler for builders.
The cloud supplier is introducing serverless mannequin customization in SageMaker, which permits builders to start out constructing a mannequin with no need to consider compute sources or infrastructure, based on Ankur Mehrotra, normal supervisor of AI platforms at AWS, in an interview with TechCrunch.
To entry these serverless model-building capabilities, builders can both comply with a self-guided point-and-click path or an agent-led expertise the place they will immediate SageMaker utilizing pure language. The agent-led function is launching in preview.
“When you’re a healthcare buyer and also you wished a mannequin to have the ability to perceive sure medical terminology higher, you’ll be able to merely level SageMaker AI, if in case you have labeled information, then choose the approach after which off SageMaker goes, and [it] fantastic tunes the mannequin,” Mehrotra stated.
This functionality is obtainable for customizing Amazon’s personal Nova fashions and sure open supply fashions (these with publicly obtainable mannequin weights), together with DeepSeek and Meta’s Llama.
AWS can also be launching Reinforcement Fantastic-Tuning in Bedrock that permits builders to decide on both a reward operate or a pre-set workflow, and Bedrock will run a mannequin customization course of robotically from begin to end.
Techcrunch occasion
San Francisco
|
October 13-15, 2026
Frontier LLMs — that means essentially the most superior AI fashions — and mannequin customization look like an space of focus for AWS at this 12 months’s convention.
AWS introduced Nova Forge, a service the place AWS will construct customized Nova fashions for its enterprise clients for $100,000 a 12 months, throughout AWS CEO Matt Garman’s keynote on Tuesday.
“Plenty of our clients are asking, ‘If my competitor has entry to the identical mannequin, how do I differentiate myself?’” Mehrotra stated. “‘How do I construct distinctive options which are optimized, that optimize my model, for my information, for my use case, and the way do I differentiate myself?’ What we’ve discovered is that, the important thing to fixing that drawback is with the ability to create custom-made fashions.”
AWS has but to realize a considerable person base for its AI fashions. A July survey from Menlo Ventures discovered that enterprises vastly choose Anthropic, OpenAI, and Gemini to different fashions. Nonetheless, the power to customise and fine-tune these LLMs may begin to give AWS a aggressive benefit.
Observe together with all of TechCrunch’s protection of the annual enterprise tech occasion right here, and see all of the bulletins you might have missed up to now right here.
