DreamhubAI/Nova-e-mini

Hugging Face
TEXT GENERATIONConcurrent Unit Cost:1Model Size:1BQuant:BF16Context Size:32kPublished:Dec 3, 2025License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Warm

DreamhubAI/Nova-e-mini, also known as WALL•E, is a 1 billion parameter language model fine-tuned from Google's Gemma 3 1B. Developed by sinamsv0, it is optimized for local, privacy-preserving AI usage in low-resource environments. This lightweight model excels at practical tasks such as text summarization, coding assistance, Linux command explanations, and providing short factual answers, with multilingual support for English, Persian, and German.

Loading preview...

WALL•E: A Lightweight Local AI Assistant

WALL•E is a 1 billion parameter language model, fine-tuned from Google's Gemma 3 1B, designed for efficient local execution and privacy-preserving AI. Developed by sinamsv0, this model prioritizes practical utility and fast responses over sheer size, making it suitable for low-resource environments.

Key Capabilities

  • Local Operation: Runs entirely on-device without requiring API keys or cloud services.
  • Multilingual Support: Offers fluent responses in English, Persian (فارسی), and German (Deutsch).
  • Practical Task Focus: Optimized for real-world applications including:
    • Text summarization (articles, notes, reports)
    • Coding assistance (Python, JavaScript, Bash, shell)
    • Linux command explanations and troubleshooting
    • Providing short factual answers and guidance
  • Efficient Prompt Handling: Designed to naturally process short and minimal prompts, avoiding over-generation.

Technical Overview

Built using Supervised Fine-Tuning (SFT) on custom multilingual datasets with safety filtering, WALL•E leverages the Unsloth framework. It operates with BF16 precision and has a context length of 3200 tokens.

Ideal Use Cases

  • Local coding assistant
  • Study and document summarization
  • Privacy-focused users
  • Lightweight edge deployments
  • Research and experimentation with small LLMs

Limitations

Due to its 1 billion parameter size, WALL•E has limited reasoning depth and is not intended for complex multi-step reasoning or creative writing tasks. It is trained with safety-aware data to avoid harmful or unethical requests.

Finetunes
1 models