OpenxAI x Ultralytics

[Cappybara12]

Hey everyone 👋

I'm Akshay from OpenxAI Labs in New York. We're building Own 1 a dedicated AI computer for running everything locally, including computer vision.

Own 1 comes with a camera integration layer and we want YOLO to be the default vision framework on it real-time object detection, scene understanding, all running on-device with zero cloud dependency.

Would love to connect with Glenn or anyone on the team about running YOLO natively on our hardware. Who's the right person to talk to?

[UltralyticsAssistant]

👋 Hello @Cappybara12, thank you for your interest in Ultralytics 🚀 and for sharing what OpenxAI Labs is building with Own 1 — exciting vision for fully local, on-device AI 👀

This is an automated response to help get you started, and an Ultralytics engineer will also assist here soon.

We recommend a visit to the Docs for new users where you can find many Python and CLI usage examples and where many of the most common questions may already be answered.

If this is a 🐛 Bug Report, please provide a minimum reproducible example to help us debug it.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset image examples and training logs, and verify you are following our Tips for Best Training Results.

For hardware, deployment, or integration discussions like this, it may help to share a bit more detail about your target runtime, OS, accelerator stack, and preferred inference path so the team can route your request more effectively.

Join the Ultralytics community where it suits you best. For real-time chat, head to Discord 🎧. Prefer in-depth discussions? Check out Discourse. Or dive into threads on our Subreddit to share knowledge with the community.

Upgrade

Upgrade to the latest ultralytics package including all requirements in a Python>=3.8 environment with PyTorch>=1.8 to verify your issue is not already resolved in the latest version:

pip install -U ultralytics

Environments

YOLO may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all Ultralytics CI tests are currently passing. CI tests verify correct operation of all YOLO Modes and Tasks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

[glenn-jocher]

Thanks for the interest, but we don’t coordinate hardware partnerships through private outreach here; please share your target OS, accelerator stack, and preferred inference path in this discussion so the team can advise publicly, and use the Export mode docs to check the supported deployment formats that best match your device.

[Cappybara12]

Thanks for the direction! Happy to share the details publicly. We're building Own 1 — a dedicated local AI computer running NixOS (our custom Linux-based OS). Here's our current stack: - OS: NixOS (custom Linux) - Accelerator: Unified memory architecture (ARM64-based, finalizing chip details) - Inference path: Ollama + llama.cpp as the primary LLM layer; looking at ONNX and TensorRT for vision workloads - Preferred export format: ONNX for cross-compatibility, open to TensorRT if we go NVIDIA GPU - Use case: Real-time object detection via camera integration on the device — detection, scene understanding, and gesture recognition all running locally Would ONNX be the recommended path for a custom Linux ARM64 device, or would TFLite make more sense given the edge constraints? Happy to go wherever the community advises.

…

Message ID: ***@***.*** com>

[glenn-jocher]

We don’t do private integration discussions here, but technically for a custom Linux ARM64 device I’d start with ONNX deployment for the best portability, move to TensorRT only if you finalize on NVIDIA, and use TFLite export mainly when your runtime is already optimized for mobile/embedded inference; for new projects, YOLO26 is the recommended baseline to benchmark first.

[Cappybara12]

Thanks @glenn-jocher for the clear guidance on ONNX/TensorRT/TFLite super helpful. We'll start benchmarking YOLO26 with ONNX on our NixOS ARM64 setup.
To give a bit more context on our side: We're building Own 1, a dedicated local AI hardware box (Intel Xe iGPU, 96 GB RAM, 4 TB SSD, 145W silent, NixOS) optimized for running agents and vision workloads completely on-device with zero cloud.
We're actively partnering with open-source communities (recently onboarded Moonshine ASR and MemOS) to co-host hackathons where we handle marketing, logistics, and prizes. A collaboration with Ultralytics/YOLO for real-time vision on Own 1 would be amazing — pre-configured support, joint hackathon track, etc.
Would love to move this to a more formal conversation if there's interest and take it up over a mail . Who's the best person on the team for hardware/ecosystem partnerships?
Happy to share more specs or run specific benchmarks.

[glenn-jocher]

Thanks for the added context — we can’t coordinate private collaboration, hackathon, or email-based partnership requests through GitHub, but if OpenxAI wants to be considered for a formal hardware/ecosystem relationship please use the Ultralytics Partner program, and feel free to keep any YOLO integration questions here in Discussions so the team can review them publicly.