Download and Manage Translation Models

MAITO’s local translation uses AI models stored on your computer. This guide shows you how to download, manage, and remove these models.

Understanding Models

What Are Translation Models?

Translation models are AI files that power MAITO’s local translation:

• File Format: ONNX (Open Neural Network Exchange)
• Size: 2-8GB per model
• Storage: %APPDATA%\Pangaia Software\MAITO\Models
• One-time Download: Download once, use forever offline

Available Model Families

Rosetta Models

• Optimized by Pangaia Software
• Trained specifically for translation
• 32 languages supported
• 2.5GB size
• Fastest inference speed

Accessing Model Management

Open Settings → Translation tab → Local AI Translation section

You’ll see five areas:

1. Installed Models - Models already on your device
2. Model Benchmarking - Test the performance of your models
3. Storage Information - Disk space usage
4. Available Models - Models available for you to download
5. Local Translation Settings - Settings to tweak the inner workings of the local translation engine

Downloading Models

Viewing Available Models

The Available Models list shows:

• Model name
• Size
• Type (CPU/GPU optimized)
• Compatibility with your hardware
• Recommendation status (⭐ for recommended)

Rosetta models appear first in the list.

Download a Model

1. Browse the Available Models list
2. Find a model that interests you (Rosetta recommended)
3. Click the Download button next to it
4. Monitor download progress
5. Wait for completion

Download Progress Shows:

• Percentage complete
• MB downloaded / Total MB
• Estimated time remaining
• Cancel button

Canceling Downloads

If you need to cancel:

1. Click Cancel button during download
2. Partial files are automatically cleaned up
3. You can restart the download later

Setting Default Model

After downloading:

1. Go to Installed Models section
2. Find and select your newly downloaded model
3. Click Set as Default
4. This model will be used for all translations

Note that the selected model in the screenshot is a sample, you may have downloaden another variant.

Managing Downloaded Models

Viewing Available Models

The Available Models section shows:

• Model name
• File size
• Download date
• Status (Active/Inactive)
• Actions (Remove, Benchmark, Set as Default)

The currently active model is highlighted.

Switching Between Models

To switch which model MAITO uses:

1. Click Set as Default on the desired model
2. MAITO reinitializes with the new model (2-3 seconds)
3. Next translation uses the new model

Removing Models

To free up disk space:

1. Locate model in Available Models
2. Ensure it’s not the active model (switch to another first)
3. Click Remove button
4. Confirm deletion
5. Model files are permanently deleted

You Cannot Remove:

• The currently active model
• Models that are downloading

Storage Information

The Storage Information section displays:

Three-Segment Breakdown:

• Models: Space used by all downloaded models
• Other: Cache, logs, temporary files
• Available: Free space remaining on drive

Visual Progress Bar:

• Blue: Models
• Gray: Other (non-MAITO) data on your disk
• Green: Available space

Model Recommendations

For CPU-Only Systems

Recommended: Rosetta 4B CPU

• Optimized for CPU inference
• Fast on modern processors
• Works on any x64 Windows system
• 2.5GB download

For GPU Systems

Recommended: Rosetta 4B DML

• DirectML GPU acceleration
• Faster than CPU version
• Requires compatible GPU
• 2.5GB download

For Low-End Systems

If your device assessment showed “Poor” or “Fair”:

Recommended: Rosetta 4B CPU

• Best performance on limited hardware
• Smallest memory footprint
• MAITO auto-adjusts context window
• Still works, just slower

Troubleshooting

”Download Failed” Error

Solutions:

1. Check internet connection
2. Verify sufficient disk space (need 2x model size temporarily)
3. Disable VPN temporarily
4. Check firewall isn’t blocking MAITO
5. Retry download (it resumes from where it left off)

“Cannot Remove Model” Error

Cause: Trying to remove the active model.

Solution:

1. Download/activate a different model first
2. Then remove the unwanted model

Model Downloaded But Won’t Load

Solutions:

1. Check model appears in Installed Models
2. Verify file size matches expected size (not truncated)
3. Remove and re-download the model
4. Check anti-virus didn’t quarantine files
5. Ensure sufficient RAM available

Very Large “Other” Storage

Cause: Cached data, logs, temporary files accumulated.

Solutions:

1. Settings → Advanced → Clear Cache
2. Manually delete logs: %APPDATA%\Pangaia Software\MAITO\logs
3. Restart MAITO

Best Practices

Download Strategy

1. Start with Recommended: Download MAITO’s recommended model first
2. Test Performance: Benchmark it (see Benchmark Guide)
3. Try Alternatives: If unsatisfied, try other models
4. Keep Favorites: Remove models you don’t use

Storage Management

• Monitor Space: Check storage info regularly
• Remove Unused: Don’t hoard models you never use
• Keep Active Models: Maintain 2-3 models for different use cases
• Plan for Updates: Models may be updated, need space for new versions

Model Selection

Prioritize:

1. Speed: If translation speed matters most → Rosetta DML
2. Compatibility: If CPU-only system → Rosetta CPU
3. Experimentation: If curious → Download multiple, compare
4. Quality: All Rosetta models have excellent quality

Advanced: Manual Model Installation

For advanced users, models can be manually placed:

Location: %APPDATA%\Pangaia Software\MAITO\Models

Structure:

Models/
  ├── Rosetta-4B-CPU-INT4/
  │   ├── model.onnx
  │   ├── model.onnx.data
  │   └── config.json
  └── Rosetta-4B-DML-INT4/
      ├── model.onnx
      └── ...

After manual placement:

1. Restart MAITO
2. Model appears in Installed Models
3. Set as default to use

What’s Next?

After downloading models:

• Benchmark Performance - Test your models
• Manage Disk Space - Free up space
• Model Selection Guide - Detailed comparison
• Understanding Local Translation - How it works

Need Help?

• Contact support: contact@pangaia.software
• Check Troubleshooting Guide