Agentic AI Comparison:
Apple Ferret-UI vs Suna by Kortix AI
vs
Introduction
This report compares Apple Ferret-UI, a research-oriented multimodal LLM specialized in mobile UI understanding and interaction, with Suna by Kortix AI, a practical AI-powered virtual assistant for business automation tasks.
Overview
Apple Ferret-UI
Ferret-UI is Apple's multimodal large language model (MLLM) designed for grounded mobile UI understanding, excelling in referring (identifying UI elements), grounding (locating elements spatially), and reasoning about screen functions. It supports any-resolution screens, outperforms GPT-4V on basic UI tasks, and is open-sourced via GitHub for research, targeting applications like smarter Siri, accessibility, and UI testing.
Suna by Kortix AI
Suna by Kortix AI is a virtual worker and AI agent platform that automates business workflows, functioning as a deployable assistant for tasks like data processing and operations. It is positioned as a practical tool with GitHub availability, emphasizing real-world business utility over specialized UI tasks.
Metrics Comparison
autonomy
Apple Ferret-UI: 9
High autonomy in UI reasoning, grounding, and executing open-ended instructions on mobile screens without external dependencies, simulating human-like navigation.
Suna by Kortix AI: 8
Strong autonomy as a virtual worker for business automation, handling independent task execution in operational contexts.
Ferret-UI edges out due to specialized agentic UI interaction capabilities, while Suna focuses on broader business autonomy.
ease of use
Apple Ferret-UI: 5
Research model requiring technical setup via GitHub and Apple ML expertise; not user-friendly for non-developers, limited to UI-specific tasks.
Suna by Kortix AI: 8
Designed as a deployable virtual assistant for business users, implying simpler integration for practical automation workflows.
Suna is more accessible for end-users, whereas Ferret-UI demands developer knowledge.
flexibility
Apple Ferret-UI: 7
Flexible for any-resolution mobile UIs across iOS/Android, with strong referring/grounding/reasoning, but narrowly focused on screen understanding.
Suna by Kortix AI: 8
Broader applicability as a general virtual worker for diverse business tasks beyond UI-specific interactions.
Suna offers wider task flexibility; Ferret-UI excels in specialized mobile UI adaptability.
cost
Apple Ferret-UI: 10
Fully open-source research model available on GitHub, free to use and deploy on compatible hardware with no licensing fees.
Suna by Kortix AI: 7
Open-source on GitHub, but as a business-oriented platform, may involve hosting/inference costs or premium features not detailed in sources.
Both open-source, but Ferret-UI's pure research nature ensures zero-cost access.
popularity
Apple Ferret-UI: 8
High research buzz with publications, GitHub repo, and media coverage outperforming GPT-4V; strong academic interest but limited commercial adoption.
Suna by Kortix AI: 5
Listed in AI agent directories with GitHub presence, but minimal mentions or detailed coverage indicate lower visibility.
Ferret-UI dominates in popularity due to Apple's backing and UI innovation hype.
Conclusions
Apple Ferret-UI leads in autonomy, cost, and popularity for specialized mobile UI agent tasks, ideal for research and on-device applications, while Suna provides superior ease of use and general flexibility for business automation. Selection depends on use case: UI navigation favors Ferret-UI; operational workflows favor Suna.