ZERO-POINT-AI/Miss_MARTHA-9B-Qwen3.5-Omni
The Miss_MARTHA-9B-Qwen3.5-Omni model by ZERO-POINT-AI is a 9 billion parameter vision-language AI, fine-tuned from Qwen3.5-9B, capable of image-text-to-text generation. Developed by Zero Point Intelligence, this model supports multimodal inputs, allowing it to see and describe images, answer questions, and perform tasks like code generation and translation. It is specifically designed for local deployment on personal hardware, emphasizing user privacy and adult companionship without data harvesting or subscriptions.
Loading preview...
Miss_MARTHA-9B-Qwen3.5-Omni: A Local Vision-Language AI
Miss_MARTHA-9B-Qwen3.5-Omni is a 9 billion parameter multimodal AI developed by Zero Point Intelligence in Dundee, Scotland. Fine-tuned from Qwen3.5-9B, this model is designed for local execution on personal computers, offering a private and subscription-free AI experience. It supports image-text-to-text generation, meaning it can process both visual and textual inputs to generate comprehensive text outputs.
Key Capabilities
- Vision and Language Understanding: Can interpret and describe images (photos, screenshots, diagrams) alongside text.
- General AI Tasks: Capable of answering questions, writing code, solving mathematical problems, and translating languages.
- Local Operation: Runs entirely on your own hardware using tools like Ollama, ensuring privacy and eliminating cloud dependencies.
- Adult-Oriented Content: Explicitly designed for adult companionship and entertainment, with a strict, hostile refusal policy for any content involving minors.
- Quantized Formats: Available in various GGUF quantizations (Q8_0, Q6_K, Q5_K_M, Q4_K_M, Q3_0, Q2_0) to accommodate different VRAM capacities, from 4GB to 8GB+.
Technical Details & Unique Features
- Base Model: Qwen3.5-9B by Alibaba Cloud.
- Architecture: Vision-Language with a 24-block ViT encoder and merger projection for vision capabilities.
- Ghost Pass: Incorporates 1e-6 noise on floating-point tensors, resulting in mathematically unique weights.
- License: Apache 2.0, allowing for broad commercial and personal use with attribution.
Good For
- Users seeking a private, locally runnable AI companion for adult-oriented conversations.
- Developers and enthusiasts who need a multimodal model for image understanding, code generation, and general text tasks without cloud reliance.
- Individuals with varying hardware specifications, as it offers optimized GGUF versions for different VRAM levels.