Numenta + Intel achieve 123x inference performance improvement for BERT Transformers
Numenta technologies combined with the new Advanced Matrix Extensions (Intel AMX) in the 4th Gen Intel Xeon Scalable processors yield breakthrough results.
Private Beta Program
Achieve dramatic performance improvements in your deep learning networks
About the Beta Program
Numenta is currently in beta with its products. For this limited closed Beta Program, we are working with a handful of customers to create highly performant, cost-effective deep learning networks for Natural Language Processing, Generative AI, and Computer Vision.
Member Benefits
- Early access to Numenta technology
- Ability to influence and shape our product
- Direct access to AI expertise backed by decades of research
- Opportunities for joint publicity
- Guidance and support for first-time deep learning deployments
- Dramatic performance improvements for Transformer and convolutional networks
- Significant cost savings
Who Should Apply
Whether you’re looking to deploy deep learning solutions for the first time or already running models in production, we look forward to working with you. Apply through the form today.
Questions?
Read the FAQ below.
Fill out my online form.
Solutions for real-time NLP and Computer Vision
<3 ms
ultra-low latency on Bert-Base
10-100x
throughput inference speedup
8x
higher throughput on Bert-Large
>50%
cost reduction on inference
10-30x
energy savings
- <3 ms ultra low latency on BERT-base
- 10-100x throughput inference speedup on CPUs or GPUs
- 8x higher throughput on Bert-Large Transformer network
- >50% cost reduction
- 10-30x energy savings
The beta is open to companies of all sizes and locations, and our customers typically fall into three categories:
- Wants to use deep learning but hasn’t yet
- Uses deep learning in production but not Transformer networks
- Uses Transformers in production
Whether you’ve never run deep learning solutions in production or are looking to accelerate multiple Transformer models, whichever path you choose, we can help you achieve significant performance improvements.
For more details, read our blog.
We focus on deploying and accelerating deep learning networks on inference, and have applied our solutions across different use cases in natural language processing and computer vision. You’ll get the most out of our beta program when:
- You are encumbered by legacy systems and traditional machine learning techniques, making it difficult to scale to new customers. We’ll guide you to new models that offer better accuracies than standard techniques, and give you the support you need to jump-start your deep learning journey.
- You would like to squeeze more out of your deep learning budget. Increase throughput and run more inferences without increasing your budget. Our products deliver dramatic performance boosts that enable larger models at lower costs.
- You are ready to enable new deep learning applications with improved accuracy. Our beta program offers a cost-effective and optimal solution to start using transformers in your applications. Our technology accelerates transformers by mapping the structures of the brain to today’s hardware, enabling efficient execution on today’s CPUs.
- You face high latencies in your real-time operating systems. For many time-sensitive AI applications, milliseconds matter. With Numenta, meet and excel in industry requirements with high throughput and ultra-low latencies.
Learn more about our solutions here.
Underlying our AI products and solutions are two decades of deep neuroscience research and breakthrough advances in cortical theory. Our research has uncovered a number of core principles of the neocortex that are not reflected in today’s machine learning systems, allowing us to define new architectures, data structures and algorithms that deliver significant benefits in today’s deep learning networks and unlock a new wave of intelligent neuroscience-based computing.
Learn more about technology behind our AI products here.
Send us a message here.
Our Solutions in Action
Turbo-charging CPU inference throughput with optimized transformer networks
Numenta Transformer models significantly accelerate CPU inference while maintaining competitive accuracy.
Breaking through the latency barrier in Conversational AI
With unique acceleration techniques built on neuroscience insights, our AI platform delivers high throughput at target low latencies for inference on CPUs.