Thanks to our customers, Quadric has outgrown its headquarters. Effective June 18, our new address is 270 East Lane, Burlingame, CA 94010.

350+ AI Models.Validated. Transparent.

No tricks, no substitutions. Every benchmark runs the complete network graph with validated accuracy. See the source code, intermediate representations, and full results.

350+

Curated Models

Fully validated with accuracy metrics

6

Model Categories

CNNs, ViTs, LLMs, Detectors, Segmenters & more

100%

Source Access

Full end-to-end graphs, no substitutions

Sample Network Benchmarks

Click on a network to see detailed performance metrics and cycle counts.

Multicore

Network FPS Cores
ResNet-50 1,330 FPS 1

Image Classification

Network QC-U
MobileNetV2-Asymmetric 4,662 FPS

Transformers Broke Your Legacy NPU

Most NPU accelerators were optimized for ResNet-style networks—simple, repetitive convolutions stacked in sequence. Modern AI looks completely different.

What Legacy NPUs Were Built For

+12 more layers

ResNets (2018-2020)

Simple, repetitive 3×3 convolutions. Fixed dataflow patterns.

What Modern AI Requires

Transformers (2023+)

Multi-head attention, LayerNorm, MatMul. Complex branching graphs.

ResNets to Transformers. Transformers to what's next.AI architectures keep evolving.

Legacy NPUs

Fixed architectures optimized for yesterday's models. Break with every major AI shift.

Chimera GPNPU

Fully programmable. Runs any operator, any graph. Ready for whatever comes next.

Continuously Expanding Coverage

Our benchmark library grows every quarter with new models across all categories—from classic CNNs to cutting-edge vision transformers and language models.

Validated Accuracy

Every model tested against reference implementations

Full Source Code

Access intermediate code and complete network graphs

Quarterly Updates

New models added as the AI landscape evolves.