ixian-bench is ZETIC's benchmarking platform for robot AI models such as π0.5, GR00T N1.7 and SmolVLA running on NVIDIA Jetson boards. It reports latency, memory and energy per inference, shows which hardware setups meet latency targets, and displays ZETIC-optimized results including an 8.1× speedup (π0.5 on Jetson AGX Thor from 254 ms to 31 ms per policy call with 78% less energy and unchanged 99% task success). Users can request benchmarks and optimizations for unlisted models.
ixian-bench is ZETIC’s web-based benchmark explorer for physical-AI models running on NVIDIA Jetson hardware. Its model library and comparison views present recorded measurements such as latency, peak board memory, energy per inference, and—in some cases—task-success results. Users can inspect benchmark conditions and methodology, compare original and ZETIC-optimized builds, and request evaluation of their own or fine-tuned models on a target board. The site cautions that task-success setups can differ across models, so those figures are not a model ranking.
Be the first Cosign
Add it to your Products I use list. You can also share why you use it.
