Skip to content

ZETIC Benchmarks

Find the right model
and hardware.

Compare inference performance across models and devices. Choose the setup that fits your workload.

Physical AIComing soon

Ixian Bench

Compare models, boards, and optimization results for Physical AI.

Latency · NVIDIA Jetson devicesExplore Ixian Bench
MobileMeasured on real devices

Melange-Bench

Compare models and execution paths on real mobile devices.

Output speed · Tokens per second · Model and device contextExplore Melange-Bench

Recorded comparisons

See the benchmarks in action.

Watch baseline and ZETIC-optimized execution side by side on the same hardware.

7 demos

π0.5

VLA

Jetson AGX Thor

37.2× faster

714 ms 19.20 ms

Policy inference · p95

LingBot-VA

VLA

Jetson AGX Thor

7.2× faster

11.15 s 1.54 s

Inference per action chunk

Sonata (PTv3)

Point-cloud segmentation

Jetson AGX Thor

23.1× faster

715 ms 31 ms

End-to-end latency per frame

Isaac-0.2-2B + SAM2

Vision & tracking

Jetson Thor

10.5× faster · peak

3.16 s 0.30 s

Prompt response

Cosmos Policy

VLA

Jetson Thor

6.7× faster

1.83 s 275 ms

Inference per 16-action chunk · p50

FoundationStereo

Stereo depth

Jetson AGX Thor

2.7× faster

104 ms 38 ms

Latency per frame

FoundationStereo · Three modes

Stereo depth

Jetson AGX Orin

3.6–8.6× faster

320 ms 37–89 ms

Inference latency per frame

Looking for Orin, Nano, or precision-specific benchmarks?

Explore Ixian BenchComing soon