Early Access Program
Be among the first teams whose robots think locally
VXI-IR5 early access is open now on our FPGA engineering evaluation platform. Tell us your robot type, host controller and target model, and our engineers will plan an evaluation path for your project stage. Early partners also help define the production specification.
A strong fit for
- System teams building AMR, inspection, mobile manipulation and service robots
- Platform teams adding on-device LLM capability alongside an existing host controller
- Projects that need on-device inference because of power, thermal or data-residency limits
What early partners get
Co-development with our own engineers
The engineers who build the silicon and toolchain work directly on model import, quantization and system integration.
Evaluation on the FPGA engineering platform
Validate your model and data flow on the engineering platform ahead of production silicon.
A voice in the production specification
Your robot and model requirements feed into the production specification and toolchain priorities.
Reproducible evaluation reports
Every report lists test conditions and steps so your team can re-run them.
How it works
Apply
Share your robot, host controller, target model and project timeline.
Align on requirements and architecture
Our technical team reviews the model, data flow and integration interfaces with you and scopes the evaluation.
Engineering-platform evaluation
Run your model on the FPGA engineering platform and receive a reproducible evaluation report.
Production planning
Plan system integration, production specification and supply together, based on the results.
Early Access Program application
The more complete your answers, the better we can plan evaluation resources and follow-up.
- Robot type：AMR / mobile robot、Robot arm / cobot、Humanoid robot、Inspection robot、Service robot、Other
- Host controller：Arm-based SoC、x86 industrial PC、GPU compute module、RISC-V / MCU、In-house or custom SoC、Not decided yet
- Model type：Large language model (LLM)、Vision-language model (VLM)、Vision-language-action model (VLA)、Speech model、Perception model (detection, segmentation, depth)、Other
- Target model
We use the information you provide only to evaluate this collaboration and follow up with you.
FAQ
- Can we apply before our model is decided?
- Yes. Choose “Not sure yet” for model type or size, and our technical team will work back from your use case to the model and data flow.
- Do we need to replace our robot's host controller?
- No. VXI-IR5 works as a co-processor next to your controller: the controller keeps motion control and safety, and VXI runs the large-model inference.
- How are evaluations prioritized after we apply?
- We plan evaluation resources based on project stage, rollout timeline and model requirements, and our technical team confirms next steps with you.
Not ready to apply yet?
Talk to our architecture team first, or use the estimator to see the memory and bandwidth your model needs.