| license: apache-2.0 | |
| language: | |
| - en | |
| - es | |
| - zh | |
| tags: | |
| - mlx | |
| - vision | |
| - vlm | |
| - osirisbrain | |
| - apple-silicon | |
| - qwen2.5-vl | |
| base_model: Qwen/Qwen2.5-VL-3B-Instruct | |
| pipeline_tag: image-text-to-text | |
| library_name: mlx | |
| # OsirisHippocampus-Vision-v7-MLX | |
| **The Hippocampus** — Osiris's visual cortex. A lightweight 3B VLM that processes screenshots, images, and visual input. Runs natively on Apple Silicon via MLX Metal. | |
| ## Architecture | |
| - **Base Model:** Qwen2.5-VL-3B-Instruct (3B parameters, vision-language) | |
| - **Format:** MLX 4-bit quantized (Apple Silicon native) | |
| - **Size:** ~2.9 GB | |
| - **Speed:** ~150+ tokens/sec on M2 Pro | |
| - **Capabilities:** OCR, screenshot analysis, image understanding, visual QA | |
| ## Usage | |
| ```python | |
| from mlx_vlm import load, generate | |
| model, processor = load("osirisbrain/OsirisHippocampus-Vision-v7-MLX") | |
| output = generate(model, processor, "What do you see in this image?", ["screenshot.png"]) | |
| ``` | |
| ## Credits | |
| MLX conversion by [mlx-community](https://huggingface.co/mlx-community/Qwen2.5-VL-3B-Instruct-4bit). | |
| Original model: [Qwen/Qwen2.5-VL-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-VL-3B-Instruct) by Alibaba. | |