MacPaw AI Models

Updated April 20, 2026
ModelEney ComponentPurpose
elix-baseLocal Engine (on-device LLM), fine-tuning of Qwen3-4B; Copyright 2024 Alibaba Cloud)Central model responsible for processing user inputs and executing actions.
elix-tool-calling-loraLoRA Adapter (Low-Rank Adaptation), running on-device.LoRA adapter running locally on the Apple MLX framework for task-specific optimization, performing Eney-style tool calling.
elix-relevance-ranker-loraLoRA Adapter (Low-Rank Adaptation), running on-device.LoRA adapter running locally on the Apple MLX framework for task-specific optimization, selecting best tools by user query and ranking them.
elix-memo-classifierLoRA Adapter (Low-Rank Adaptation), running on-device.Apple MLX framework for task-specific optimization, classifying whether a user query is a part of long-term memory.
elix-query-guardrailLoRA Adapter (Low-Rank Adaptation), running on-device.LoRA adapter running locally on the Apple MLX framework for filtering abusive content.
mnemos_provider_classifier-v1.5Local Engine (machine learning model)Dynamically sourced local machine learning model, used for input classification.
bge-small-en-v1.5-bf16Local Engine (machine learning model)An embedding model component used to process and organize data for fast, meaning-based search capabilities within the application.

Can’t find the answer to your question?

Contact us and we’ll get back to you as soon as we can.

Contact Us