Skip to content

v3.0.1

Latest
Latest

Choose a tag to compare

View all tags
@smshagor-dev smshagor-dev released this 05 Sep 10:16
v3.0.1
1d740ce

OpenMindAI v3.0.1

OpenMindAI v3 introduces the next major stage of the OpenMindAI ecosystem, expanding from a desktop-focused local AI application into a cross-platform AI platform for Windows, macOS, Linux, Android, and iOS.

Version 3.0.0 introduced the major new architecture, mobile applications, expanded local AI capabilities, model management, multimodal tools, document workflows, and automatic model routing.

Version 3.0.1 focuses on stability, model-routing reliability, mobile chat fixes, model download experience, and general usability improvements.

OpenMindAI v3.0.0

OpenMindAI Mobile

  • Introduced the OpenMindAI Mobile application.
  • Added Android support.
  • Added iOS application architecture.
  • Added a mobile-first ChatGPT-style chat interface.
  • Added local AI model support for compatible mobile devices.
  • Added mobile model installation and management.
  • Added automatic model selection for different AI tasks.
  • Added model capability information directly inside the mobile application.
  • Added dedicated interfaces for chat, reasoning, image, and document workflows.
  • Unified the overall experience between OpenMindAI Desktop and OpenMindAI Mobile.

Cross-Platform OpenMindAI Ecosystem

OpenMindAI now targets:

  • Windows
  • macOS
  • Linux
  • Android
  • iOS

The goal of the v3 architecture is to provide the same OpenMindAI experience across desktop and mobile while allowing each platform to use hardware-appropriate local AI models.

Local and Offline AI

  • Expanded local AI inference support.
  • AI models can run directly on the user's device.
  • Local conversations can continue without an active internet connection after required models and runtimes are installed.
  • Reduced dependency on external AI APIs.
  • Local-first architecture provides better privacy for supported workflows.
  • Models, conversations, documents, and generated content can remain on the user's own device.
  • Added hardware-aware model recommendations.
  • Improved support for lightweight models for lower-resource devices.
  • Added support for larger models on capable desktop hardware.

New Model System

The model system received a major redesign.

OpenMindAI now uses simplified OpenMindAI model names instead of exposing complicated provider, architecture, and quantization information throughout the normal user interface.

Model families include:

  • OpenMindAI Nano
  • OpenMindAI Swift
  • OpenMindAI Core
  • OpenMindAI Titan
  • OpenMindAI Reasoning Mini
  • OpenMindAI Reasoning
  • OpenMindAI Lens

Different models are optimized for different hardware and workloads.

Expanded Open Model Support

The v3 model architecture was expanded to support additional openly available model families and compatible local AI models.

The underlying ecosystem can use models originating from open model families including technologies from:

  • OpenAI
  • Google
  • Meta
  • Alibaba and Qwen ecosystem
  • other compatible open-weight model providers

Provider-specific technical names are abstracted where appropriate so users can select models based primarily on capability instead of model implementation details.

Model Capability System

Models can now expose the tasks they are designed to perform.

Capabilities may include:

  • Chat
  • Writing
  • Coding
  • Reasoning
  • Deep Thinking
  • Research
  • Vision
  • Image Understanding
  • Document Analysis
  • Tool Use
  • Agentic Workflows

This makes it easier to understand which model should be used before downloading or activating it.

Automatic Model Routing

Added an improved model-routing architecture.

OpenMindAI can select an appropriate installed model based on:

  • User request
  • Required capability
  • Model availability
  • Device resources
  • Hardware compatibility
  • Task type

This reduces the need to manually change models for every task.

Manual model selection remains available when users want direct control.

Hardware-Aware AI

OpenMindAI can use device capabilities when determining appropriate local models.

The model system is designed to consider factors such as:

  • Available RAM
  • CPU capabilities
  • GPU availability
  • Available VRAM
  • Device type
  • Operating system
  • Model size
  • Workload requirements

This helps prevent unnecessarily large models from being recommended to devices that cannot run them efficiently.

Model Manager

The local Model Manager was expanded with:

  • Model discovery
  • Model installation
  • Download management
  • Installed-model detection
  • Model removal
  • Model reinstallation
  • Recommended models
  • Capability information
  • Hardware compatibility information
  • Storage awareness
  • Local model status

Users can install additional models whenever they need more capabilities.

Recommended Model Setup

The setup workflow now provides recommended models instead of forcing users to manually determine which models should be installed.

Multiple compatible recommended models can be shown based on the device rather than displaying only a single recommendation.

A lightweight general-purpose model can be installed first so users can begin using local chat quickly and install specialized models later.

Local AI Chat

The chat system received a major interface and architecture update.

Improvements include:

  • Cleaner ChatGPT-style interface
  • Streaming AI responses
  • Model-aware conversations
  • Automatic model routing
  • Manual model selection
  • Local conversation history
  • Copy response support
  • Markdown rendering
  • Code rendering
  • Improved input area
  • Better conversation layout
  • Improved mobile responsiveness

Reasoning and Deep Thinking

OpenMindAI v3 expands reasoning-oriented workflows.

Supported model configurations can be used for:

  • Complex reasoning
  • Problem solving
  • Technical analysis
  • Programming
  • Research assistance
  • Step-based planning
  • Mathematical and scientific tasks

Dedicated reasoning models can be selected automatically when appropriate.

Coding and Development

Expanded support for software-development workflows.

OpenMindAI can assist with:

  • Code generation
  • Code explanation
  • Debugging
  • Refactoring
  • Architecture analysis
  • Documentation
  • Development planning
  • Repository-oriented workflows

The architecture also prepares OpenMindAI for more advanced coding-agent capabilities.

Vision and Image Understanding

Added and expanded multimodal AI workflows.

Supported models can analyze:

  • Images
  • Screenshots
  • UI designs
  • Diagrams
  • Documents containing visual information

Vision-capable models are represented through the OpenMindAI capability system.

Image Generation

Added a dedicated image-generation experience.

The interface separates image-generation workflows from normal text chat, making it easier to create and manage generated visual content.

Documents

Expanded document workflows.

OpenMindAI can support document-oriented tasks including:

  • Document creation
  • Document analysis
  • Summarization
  • Editing
  • Structured content generation
  • PDF workflows
  • DOCX workflows
  • Markdown workflows

Dedicated document interfaces were added to make document generation different from ordinary chat.

PDF and DOCX Generation

Added dedicated document generation workflows for creating downloadable document formats including:

  • PDF
  • DOCX
  • Markdown

This provides a more structured environment for reports, research material, professional documents, and other long-form content.

Research Workflows

Expanded OpenMindAI research capabilities.

The architecture supports workflows such as:

  • Research assistance
  • Document analysis
  • Source-oriented investigation
  • Technical research
  • Long-form synthesis
  • Reasoning-based analysis

Online research functionality can complement local models when internet-based information is required.

Voice and Audio Architecture

The v3 architecture includes support for voice-oriented AI functionality.

Supported workflows can include:

  • Voice input
  • Speech-related AI interaction
  • Audio-capable model integration

Availability depends on the platform and installed capabilities.

Projects

OpenMindAI projects provide a way to organize related:

  • Conversations
  • Documents
  • Research
  • Generated files
  • AI workflows

This creates a persistent workspace instead of treating every conversation as an isolated chat session.

Privacy

OpenMindAI continues to follow a local-first architecture.

When using local models:

  • Prompts can be processed directly on the user's device.
  • Local documents do not need to be uploaded to an external AI provider.
  • Conversations can remain locally stored.
  • Installed AI models remain under the user's control.
  • Core offline functionality does not require a continuous internet connection.

Internet access is still required for functionality such as initial model downloads, updates, web-based research, or explicitly connected online services.

OpenMindAI v3.0.1

Version 3.0.1 is a stability and usability update built on top of the major v3.0.0 architecture.

Mobile Chat Fixes

Fixed issues where a locally installed model could fail when a message was submitted from the mobile chat interface.

Improvements were made to:

  • Chat request handling
  • Installed-model detection
  • Model loading
  • Selected-model state
  • Model availability checks
  • Chat-to-model routing
  • Local inference initialization

Model Routing Improvements

Improved automatic model routing across OpenMindAI.

The routing system now handles installed models and supported capabilities more reliably.

Improvements include:

  • Better model availability detection
  • Improved fallback behavior
  • More reliable capability matching
  • Improved routing between chat and specialized models
  • Better handling when the preferred model is not installed
  • Reduced unnecessary model-selection failures

Model Download Progress

Improved the model download interface.

Model downloads can now expose clearer progress information including:

  • Download progress
  • Percentage completed
  • Download status
  • Download speed
  • Model installation state

This makes large local model downloads easier to understand and monitor.

Recommended Models Improvement

The model setup flow was improved so that compatible recommended models can be presented together rather than restricting setup to only one model.

Users can better understand which models are recommended for:

  • General chat
  • Reasoning
  • Coding
  • Vision
  • Other supported AI tasks

Improved Model Cards

Model information was simplified for normal users.

Instead of emphasizing internal information such as:

  • Original provider
  • Original model family
  • Quantization identifiers
  • Low-level implementation names

the interface focuses primarily on:

  • OpenMindAI model name
  • Model purpose
  • Supported capabilities
  • Hardware suitability
  • Download size and status
  • Installed state

Technical metadata can remain available internally where required.

Mobile UI Improvements

Improved the mobile application interface across major AI workflows.

Updates include:

  • Cleaner chat screen
  • Improved model interaction
  • Better model-management screens
  • Improved capability presentation
  • Dedicated image-generation experience
  • Dedicated document-generation experience
  • Better responsive layout
  • Improved navigation

Installed Model Detection

Improved detection of models that are already downloaded and installed.

This prevents the application from unnecessarily treating an installed model as unavailable or attempting to repeat setup steps.

Model Installation Reliability

Improved model installation state management.

The application now handles transitions between:

  • Not installed
  • Downloading
  • Installing
  • Installed
  • Loading
  • Ready
  • Failed

more consistently.

Stability Improvements

Multiple internal improvements were made across the desktop and mobile architecture.

These include:

  • Better error handling
  • Improved model-state synchronization
  • Improved local inference initialization
  • Better routing consistency
  • More reliable application startup
  • Better model-manager state handling
  • Improved cross-platform behavior
  • General UI and UX fixes

Desktop Distribution

OpenMindAI Desktop v3.0.1 includes packaged Windows x64 distribution support.

Desktop bootstrap and setup architecture also covers:

  • Windows x64
  • macOS
  • Linux x64

Platform-specific runtime and hardware capabilities may differ.

Security and Privacy Improvements

The local-first architecture continues to minimize unnecessary external transmission of user content.

v3.0.1 improves internal state handling around local models and keeps local AI functionality separated from optional connected or online functionality.

OpenMindAI Ecosystem

With v3.0.1, OpenMindAI is evolving into a unified AI environment covering desktop and mobile devices.

The ecosystem now combines:

  • Local AI
  • Offline Chat
  • Automatic Model Routing
  • Model Manager
  • Reasoning
  • Coding
  • Vision
  • Image Generation
  • Documents
  • PDF and DOCX Generation
  • Research
  • Voice architecture
  • Projects
  • Hardware-Aware AI
  • Mobile AI
  • Desktop AI
  • Privacy-Focused Local Processing

Supported Platforms

  • Windows
  • macOS
  • Linux
  • Android
  • iOS

Upgrade

Users running OpenMindAI v3.0.0 are recommended to upgrade to v3.0.1 for improved model routing, mobile chat reliability, model installation handling, download progress reporting, and general stability improvements.

Thank You

OpenMindAI is being built as an open, practical, privacy-focused AI platform that allows users to run intelligent models directly on their own hardware while still providing access to more advanced connected capabilities when needed.

More improvements are already being developed for future OpenMindAI releases.

Website: https://www.openmindai.org/

Release: OpenMindAI v3.0.1
Previous Major Release: OpenMindAI v3.0.0
Platforms: Windows, macOS, Linux, Android, iOS

Assets 21
Loading