Introducing TokenProcessor™

Building The Future Of AI Compute

A new approach to AI acceleration that prioritizes adaptability alongside performance, enabling long-term value across evolving workloads and product lifecycles.

Request Product Brief

TOPS (INT8)

5–10W

Power Envelope

64GB

LPDDR4

VLM

COMPATIBLE

Backed by Leading Investors

The Challenge

AI Evolves. Hardware Doesn’t.

AI models, algorithms, and workloads are changing faster than ever. Yet many AI accelerators remain optimized for a narrow set of architectures, creating a growing gap between what future applications require and what existing hardware can support.

AI Is Moving Fast

CNNs

2010s

Transformers

2017+

Large Language Models

2020+

Multimodal AI

2023+

Physical AI & Beyond

AI Never Stops Evolving

New models and architectures emerge every year - opening up new possibilities and new demands.

Hardware Falls Behind

AI Capability & Complexity

Hardware Flexibility

2010

2015

2020

2025

2030+

ꜜ

↑

⚠

The Gap

The Gap Keeps Growing

Hardware optimized for today’s models struggles to keep up with tomorrow’s workloads.

The Cost of Inflexibility

Specialized Architectures

Built for specific models, not general workloads

Shorter Useful Life

Hardware becomes obsolete before the product lifecycle ends

Higher TCO

More redesigns, re-qualifications, and replacement cycles

Ecosystem Lock-In

Closed platforms limit flexibility and innovation

Rigid Today. Risky Tomorrow.

Inflexible hardware increases costs, slows innovation, and creates long-term risk.

The LENZO Approach

Built to Adapt as AI Evolves

Adaptability

Deployability

Long-Term Value

Adaptability

Deployability

Long-Term Value

Adaptability

Most AI accelerators are designed around a narrow set of workloads. TokenProcessor™ was built with a different philosophy: support a broad range of computational patterns so hardware can continue delivering value as AI evolves.

CNNs

→

Transformers

→

LLMs

→

Multimodal

→

Physical AI

→

Future

←

TokenProcessor™

→

Deployability

AI does not live only in data centers. Real-world systems, from robotics and industrial vision to autonomous infrastructure, require hardware that can operate efficiently across diverse environments and product lifecycles.

Robotics

Industrial Vision

Autonomous Systems

Edge Infrastructure

Long-Term Value

When AI changes faster than hardware refresh cycles, flexibility becomes a business advantage. By reducing the need for redesigns and hardware replacement, TokenProcessor™ helps organizations maximize the return on their AI investments.

Traditional

Deploy

→

AI Changes

→

Redesign

→

Repeat

TokenProcessor™

Deploy

→

AI Changes

→

Continue

→

Sustained Value

LENZO A-Series

Built For Demanding AI Workloads

The A-Series combines the adaptability of TokenProcessor™ with the performance, efficiency, and developer ecosystem required for real-world deployment.

Request Product Brief

65 TOPS INT8

High-performance AI acceleration for edge and Physical AI workloads.

5–10W Power

Efficient operation for embedded, industrial, and edge deployments.

5–10W Power

Price management helps adjust and control pricing

Up To 64GB LPDDR4

Support larger models and memory-intensive applications.

PCIe Gen3

High-bandwidth connectivity for scalable system integration.

PCIe Gen3

Smart Forecasting uses real-time data and AI to predict

Linux Environment

Built for familiar development workflows and deployment pipelines.

Framework Support

PyTorch, TensorFlow, ONNX, and other popular AI frameworks.

Use Cases

Built For The Real World

From robotics and industrial automation to autonomous systems and intelligent infrastructure, the A-Series is designed for AI workloads that interact with the physical world.

Robotics

Autonomous Systems

Advanced Perception

Robotics

Autonomous Systems

Advanced Perception

Robotics

Autonomous Systems

Advanced Perception

The Technology

Built On A Different Computing Model

TokenProcessor™ is built on LENZO’s proprietary CPU-Grounded Linear Array (CGLA) architecture. Developed through years of research at NAIST, CGLA combines the flexibility of programmable computing with the efficiency of specialized acceleration, enabling a broader range of computational patterns to be executed on a single platform.

List of Publications

View the full list of publications of the NAIST Computing Architecture Lab, including peer-reviewed, invited-talk & awards, and over 60 patents.

News

Latest From LENZO

Follow product updates, technical insights, research milestones, industry events, and company announcements.

LENZO A-Series Products Materials Now Available - In JA and EN

LENZO Featured in Nikkei Tech Foresight

LENZO Featured in Three-Part ZEROICHI Interview Series with Takafumi Hori

LENZO to Speak at Monozukuri Innovation Meet Up in Kyoto

LENZO to Present at CadenceLIVE Japan 2026

LENZO A-Series Products Materials Now Available - In JA and EN

LENZO Featured in Nikkei Tech Foresight

LENZO Featured in Three-Part ZEROICHI Interview Series with Takafumi Hori

EVENTS

LENZO to Speak at Monozukuri Innovation Meet Up in Kyoto

Request Product Brief

Let’s Build What’s Next

Whether you’re exploring robotics, advanced perception, autonomous systems, or next-generation edge infrastructure, we’d love to learn more about your application.

Let’s Build What’s Next

Whether you’re exploring robotics, advanced perception, autonomous systems, or next-generation edge infrastructure, we’d love to learn more about your application.

Let’s Build What’s Next

Whether you’re exploring robotics, advanced perception, autonomous systems, or next-generation edge infrastructure, we’d love to learn more about your application.

Get in Touch