E1Veyra E1
Compact INT8 inference for always-on audio and sensor workloads.
- INT8 precision
- 6×6 mm QFN-48
Hosted evaluation boards · E-seriesEdge intelligence,
Edge intelligence,
built for production.
High-efficiency NPUs and edge boards for always-on and vision workloads.
E-series processors
Production silicon and boards for always-on audio, sensor, and vision workloads at the edge.
E1
E2Veyra E2
Balanced NPU for vision at the edge, with FP16 and INT8 profiles.
- FP16 · INT8 precision
- 2× MIPI CSI-2
E4Veyra E4
Edge computer board for multi-stream deployments.
- FP16 · INT8 precision
- 8 GB LPDDR5
Evaluate before you commit.
Veyra AI Hub lets developers test models on exact hosted boards, then take the results home.
- 01
Bring a model
ONNX or TFLite, or a ready-to-run example.
- 02
Pick an exact board
Compatibility and availability shown before you build.
- 03
Compare & export
Traceable metrics and a local reproduction package.
Developer resources
Public guides and packaging notes for Veyra silicon.
- v2.1DownloadRead in Hub →E2 Quick StartVY-E2 Eval Rev B · SDK 2.1
- v2.1DownloadRead in Hub →E2 Quantization GuideINT8 preparation and calibration
- v1.2DownloadRead in Hub →Input GuideInput contracts for example models
- v2.1DownloadRead in Hub →Local HandoffReproduce a run package on your own bench