Bertha 100

Coming Q4, 2026
Pioneering Agentic AI with Versatile Edge Computing

Optimized for real-time processing of multi-modal AI inference

Ask Inquiry

Real-time processing and user personalization

Bertha 100 (B100) is one-of-a-kind accelerator to actualize Agentic AI. B100 is capable of encoding, inference, and diffusion of text, image, audio, and video for AI decision-making and action execution.
B100 is catered towards practicality with its balanced specification (i.e., 48 TFLOPS, 64 GB/s)  to meet performance and power requirements of IoT devices, smart appliances, and on-premise datacenters. B100 provides a solution closer to the user for real-time processing without additional network latency and user personalization while ensuring data privacy.

Key Features

LPU-based Architecture

Streamlined memory access with precise alignment of memory bandwidth and compute bandwidth for 90% hardware utilization during inference. Integration of peripheral processors to enable multi-modal computation.

Soc Integration

Advanced integration of LPU fabricated with 4nm technology node, 4 channels of LPDDR5x for state-of-the-art system-on-chip design. Offered as PCIe card or as IP to customer needs.

Multi-chip Scalability

Custom on-chip network controller for computation-communication overlapping to hide the communication overhead and achieve near-perfect scalability.

HyperDex Software

Plug & play solution for seamless serving Generative AI applications on HyperAccel hardware. Support for standardized ML frameworks for inference(e.g., PyTorch, ONNX, vLLM) with SDKs for further optimizations, deployment, and profiling based on user needs.

Specifications

FP8
32.768TFLOPS
Target Frequency
1.0GHz
Number System
BF16, FP8, FP4, INT8, INT4
DRAM Bandwidth
LPDDR5x, 64 GB/s
DRAM Size
16GB
Form factor
M.2