Model Support

EdgeMob is designed to provide developers with maximum flexibility in how they run and integrate AI models. The platform supports both open-source and custom models, ensuring that developers can experiment, deploy, and scale AI workloads without depending on closed ecosystems.

Open-Source Model Integration

• EdgeMob natively supports popular open-source models such as LLaMA, Mistral, Qwen, and CodeLLaMA, along with smaller models optimized for on-device execution.

• Developers can load these models directly into the EdgeMob app using built-in import tools or via the model registry.

• Continuous updates to EdgeMob’s runtime ensure compatibility with new model formats and community-driven improvements.

Custom Model Integration

• Developers are not limited to pre-approved models. They can package and load their own custom models into EdgeMob.

• The platform provides SDKs and documentation for preparing model weights, quantization, and deployment pipelines.

• Custom models can be exposed locally (for testing) or through the API Gateway (for broader application integration).

Extensible Runtime

• The EdgeMob orchestrator, built in Rust and C++, integrates with frameworks like Candle and llama.cpp, enabling efficient inference across multiple architectures.

• This extensibility ensures that future frameworks or model families can be adopted with minimal friction.

Model Registry & Marketplace

• EdgeMob includes a model registry where developers can publish, share, and access models.

• In future iterations, a marketplace will allow developers to monetize models, while node operators can choose which models to host based on demand.

---

By supporting both open-source and custom models, EdgeMob lowers barriers to innovation and empowers developers to build AI applications that are cost-efficient, privacy-preserving, and tailored to their unique needs.