AVAI
Mobilint
Software Development Kit

SDK qb

Full-stack software development kit for Mobilint NPUs

Supported Models
400+
INT8 Accuracy
99%
Platforms
Win + Linux
Verification
MLPerf
SDK qb

Overview

SDK qb

SDK qb is Mobilint's tailor-made development toolkit. It takes models trained on GPUs in standard frameworks and compiles them for Mobilint NPU hardware — handling quantization, optimization, and runtime execution automatically.

The workflow is simple: train on GPU using your existing pipeline, export from PyTorch, TensorFlow, ONNX, TFLite, or Keras, compile with SDK qb (auto INT8 quantization at 99% accuracy), and deploy to any Mobilint NPU.

Supports both C++ and Python APIs for integration into production systems. The same MXQ binary runs on any Mobilint NPU — from the 3W REGULUS SoC to the 80 TOPS ARIES accelerator.

Mobilint

Authorised Mobilint Distributor
Official sales, integration & support services

Features

Key Capabilities

No Retraining Required

Your GPU-trained model compiles directly. No architecture changes, no new training runs. The compiler handles quantization and optimization end-to-end.

Advanced Quantization

Maintains over 99% of the original model's accuracy after INT8 quantization. Proprietary calibration algorithms minimize precision loss.

Cross-Platform Support

Develop and compile on Windows or Ubuntu. The same MXQ binary runs on any Mobilint NPU — from REGULUS (3W) to ARIES (80 TOPS).

Production-Ready APIs

Intuitive C++ and Python APIs for quick NPU integration. Minimal boilerplate to load models, run inference, and manage multi-device deployments.

Specifications

Technical Details

Supported FrameworksPyTorch, TensorFlow, TFLite, ONNX, Keras
Output FormatMobilint Executable (MXQ)
QuantizationAutomatic INT8 (99% accuracy retention)
Runtime APIsC++ and Python
OS SupportWindows, Ubuntu
Validated Models400+ architectures
VerificationMLPerf Verified
Multi-DeviceMulti-instance runtime with dynamic workload balancing

Applications

Use Cases

Model Compilation

Compile models from any supported framework into optimized MXQ binaries for Mobilint NPU hardware.

AI Research & Dev

Rapid model prototyping and benchmarking on NPU hardware. Iterate on architecture and quantization without cloud queues.

Production Deployment

C++ and Python runtime APIs for seamless integration into production inference pipelines with multi-device scheduling.

Downloads

Documents & Datasheets

Request product documentation and our team will email you the files directly.

SDK qb Documentation

user guide

Ready to Get Started with the SDK qb?

Contact our team for pricing, technical questions, or to request an evaluation unit.