distil-labs/distil-email-classifier: Local Email Classification
This model, developed by Distil Labs, is a fine-tuned 0.8 billion parameter Qwen3-based model specifically designed for local email classification. It enables users to auto-label emails without sending sensitive content to external cloud LLMs, ensuring privacy and data sovereignty.
Key Capabilities
- 10-way Email Classification: Accurately categorizes emails into predefined labels such as Billing, Newsletter, Work, Personal, Promotional, Security, Shipping, Travel, Spam, and Other.
- High Accuracy: Achieves 93% accuracy on its classification task, matching the performance of its larger teacher model (GPT-OSS-120B) through knowledge distillation and supervised fine-tuning.
- Local Deployment: Designed to run entirely on a local machine using Ollama and n8n, preventing email content from leaving the user's environment.
- Integration with n8n: Provides pre-built n8n workflows for real-time classification of incoming emails and batch processing of existing emails.
- Customizable: Users can distill custom versions of this classifier with different label sets using the Distil Labs platform.
Good For
- Users requiring privacy-preserving email automation.
- Developers and individuals looking to implement local AI solutions for email management.
- Automating email organization and reducing manual labeling efforts.
- Integrating AI classification into existing workflows using tools like n8n.