Customer Reviews (125)

dontcareever • 2025-03-22T07:30:56-07:00

why is there no gemma3 being offered? also, would love some multi-modal options to process images…

jacksondonaghy • 2025-03-13T18:00:58-07:00

I’ll upgrade it when we can access Gemma 3

MikeW805 • 2025-03-13T14:02:54-07:00

I am glad I ran across this. Works quite well. Also what a deal!

Summary of Key Insights

  • Private LLM is a privacy-focused AI assistant that runs locally on iOS and macOS devices. It offers a selection of open-source LLM models, Siri & Shortcuts integration, and customizable interactions.
  • The app is generally well-received, with users appreciating its privacy features, offline capabilities, and model selection. However, there are concerns about stability, missing features (like chat history, document upload, and custom model import), and some limitations in AI response quality.
  • The overall sentiment is positive, but there's room for improvement in user experience and feature set to better compete with other LLM apps.
  • The overall rating, calculated from the provided score distribution, is approximately 4.05 stars. The score distribution shows a significant number of 5-star (48.8%) and 4-star (20.8%) reviews, indicating overall satisfaction. However, there is a notable percentage of 1-star (12%) and 2-star (7.2%) reviews, highlighting areas of concern.

Data & Methodology Overview

  • Data Sample Size: 125 reviews
  • Time Frame: Not explicitly stated, but assumed to be recent given the app version and model mentions.
  • Analysis Methods:
    • Sentiment Analysis: Implicitly derived from star ratings and review text.
    • Keyword Extraction: Manually identified recurring themes, feature requests, and pain points from review text.
    • Frequency Analysis: Counted mentions of specific issues and features to prioritize recommendations.

Key User Pain Points

  • Stability Issues: Several users reported crashes, especially with specific models (e.g., Qwen2.5-Coder, 7B models on certain iPhones). This significantly impacts usability.
  • Lack of Chat History/Multiple Conversations: Many users requested the ability to save and resume conversations, as well as manage multiple chats simultaneously. The absence of this feature is a major drawback.
  • Inability to Import Custom Models: Users want the flexibility to download and use models from Hugging Face or other sources. The current limitation to pre-selected models restricts customization.
  • Limited Functionality: Missing features like document upload, web search, and image processing are frequently mentioned.
  • Response Quality Issues: Some users found the AI responses to be inaccurate, nonsensical, or overly cautious, especially when compared to ChatGPT or other online LLMs. Also, some users are finding the models to be crippled by the app's settings.
  • UI/UX Annoyances: Issues with scrolling, text input, font size, and shortcut integration were reported, detracting from the user experience.
  • Resource Intensive: Some users with older devices reported performance issues and crashes due to high memory usage.

Frequently Requested Features

  • Chat History/Saved Conversations: This is the most frequently requested feature.
  • Document Upload/Processing: Users want to upload PDFs, DOCX, and TXT files for the AI to analyze and answer questions.
  • Custom Model Import (Hugging Face Integration): Allowing users to add their own models.
  • Web Search Integration: Enabling the AI to access and incorporate information from the internet.
  • Image Processing (Multimodal Capabilities): The ability to process and understand images.
  • Voice Support: Voice input and output capabilities.
  • Dark Mode: A dark theme for improved readability and reduced eye strain.
  • Code Execution: Ability for the AI to run code and incorporate the output in its responses.
  • Increased Font Size: An option to increase the font size for better readability.
  • Customizable System Prompts per Chat: Ability to set different system prompts for different conversations.

Strengths and Positive Aspects

  • Privacy: The primary selling point is the local, offline processing, ensuring user data privacy.
  • Offline Capabilities: The ability to use the AI without an internet connection is highly valued.
  • Model Selection: Users appreciate the variety of pre-selected LLM models.
  • Siri & Shortcuts Integration: Integration with the Apple ecosystem enhances accessibility.
  • Ease of Use: Many users find the app simple and straightforward to use.
  • Performance on Newer Devices: Users with newer iPhones and Macs report good performance, especially with smaller models.
  • Developer Responsiveness: Some users praised the developers' knowledge, diligence, and responsiveness to feedback.
  • One-Time Purchase: The one-time purchase model is preferred over subscription-based alternatives.

Prioritized Action Recommendations

High Priority:

  • Address Stability Issues: Investigate and fix crashes, especially with specific models and devices. Impact: Prevents usage and creates negative user experience.
  • Implement Chat History/Saved Conversations: Add the ability to save, resume, and manage multiple conversations. Impact: Improves usability and allows for more complex interactions.
  • Improve UI/UX: Fix scrolling issues, optimize text input, and add font size customization. Impact: Enhances overall user experience and accessibility.

Medium Priority:

  • Enable Document Upload/Processing: Allow users to upload documents for analysis and question answering. Impact: Expands the app's utility for research and information retrieval.
  • Explore Custom Model Import (Hugging Face Integration): Consider allowing users to add their own models, perhaps with a warning about compatibility. Impact: Increases customization and caters to advanced users.
  • Implement Web Search Integration: Enable the AI to access and incorporate information from the internet (while considering privacy implications). Impact: Improves the accuracy and relevance of AI responses.
  • Add Dark Mode: Provide a dark theme option. Impact: Improves user experience, especially in low-light environments.

Low Priority:

  • Implement Voice Support: Add voice input and output capabilities. Impact: Enhances accessibility and allows for hands-free interaction.
  • Implement Code Execution: Enable the AI to run code and incorporate the output in its responses. Impact: Useful for developers and programmers.
  • Add Image Processing (Multimodal Capabilities): The ability to process and understand images. Impact: Expands the app's utility, but may require significant resources.

Opportunities for Startup Ideas

  • Privacy-Focused AI-Powered Document Summarization and Analysis Tool: Build an app specifically designed for summarizing and analyzing local documents, ensuring complete privacy.
  • Local LLM Marketplace: Create a platform where users can download and share custom-trained LLM models for specific tasks, compatible with apps like Private LLM.
  • AI-Powered Personal Knowledge Base: Develop an app that uses local LLMs to organize and retrieve information from a user's notes, documents, and other personal data.
  • Offline AI Education Platform: Build an app that uses local LLMs to provide personalized learning experiences without requiring an internet connection.

Trends and Observations

  • Demand for Larger Models: Users with newer devices are consistently asking for support for larger, more capable LLM models.
  • Focus on Privacy: The privacy aspect is a major draw for users, suggesting a strong market for privacy-focused AI solutions.
  • Feature Parity with Online LLMs: Users are increasingly expecting offline LLM apps to offer similar functionality to online services like ChatGPT.
  • Device Compatibility Issues: There's a recurring theme of compatibility problems with older devices, highlighting the need for careful optimization and clear system requirements.

Conclusion

Private LLM offers a compelling value proposition with its privacy-focused, offline AI capabilities. However, to truly compete in the evolving LLM landscape, the app needs to address stability issues, improve the user experience, and expand its feature set. By prioritizing the recommendations outlined above, the developers can solidify Private LLM's position as a leading choice for users seeking private and versatile AI assistance on their iOS and macOS devices.

Original App Link

https://apps.apple.com/us/app/private-llm-local-ai-chat/id6448106860?uo=2